Merge branch 'master' into cleanup_deps

Drop also ttf files
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
2026-02-03 11:13:31 -05:00 · 2024-08-21 13:10:46 +02:00 · 2024-08-21 13:03:26 +02:00 · 2024-08-21 13:02:19 +02:00 · 2024-08-21 13:02:19 +02:00 · 2024-08-21 13:02:19 +02:00
182 changed files with 1802 additions and 5109 deletions
--- a/.devcontainer-scripts/utils.sh
+++ b/.devcontainer-scripts/utils.sh
@@ -9,7 +9,6 @@
 # Param 2: email
 #
 config_user() {
-    echo "Configuring git for $1 <$2>"
    local gcn=$(git config --global user.name)
    if [ -z "${gcn}" ]; then
        echo "Setting up git user / remote"
@@ -25,7 +24,6 @@ config_user() {
 # Param 2: remote url
 #
 config_remote() {
-    echo "Adding git remote and fetching $2 as $1"
    local gr=$(git remote -v | grep $1)
    if [ -z "${gr}" ]; then
        git remote add $1 $2
@@ -34,22 +32,18 @@ config_remote() {
 }

 # Setup special .ssh files
-# Prints out lines of text to make things pretty
+#
 # Param 1: bash array, filenames relative to the customization directory that should be copied to ~/.ssh
 setup_ssh() {
-    echo "starting ~/.ssh directory setup..."
-    mkdir -p "${HOME}.ssh"
-    chmod 0700 "${HOME}/.ssh"
-    echo "-----"
    local files=("$@")
-    for file in "${files[@]}" ; do
+    for file in "${files[@]}"; then
        local cfile="/devcontainer-customization/${file}"
-        local hfile="${HOME}/.ssh/${file}"
+        local hfile="~/.ssh/${file}"
        if [ ! -f "${hfile}" ]; then
-            echo "copying \"${file}\""
+            echo "copying ${file}"
            cp "${cfile}" "${hfile}"
            chmod 600 "${hfile}"
        fi
    done
-    echo "~/.ssh directory setup complete!"
+    ls ~/.ssh
 }
--- a/.github/check_and_update.py
+++ b/.github/check_and_update.py
@@ -29,14 +29,9 @@ def calculate_sha256(file_path):
 def manual_safety_check_hf(repo_id):
    scanResponse = requests.get('https://huggingface.co/api/models/' + repo_id + "/scan")
    scan = scanResponse.json()
-    # Check if 'hasUnsafeFile' exists in the response
-    if 'hasUnsafeFile' in scan:
-        if scan['hasUnsafeFile']:
-            return scan
-        else:
-            return None
-    else:
-        return None
+    if scan['hasUnsafeFile']:
+        return scan
+    return None

 download_type, repo_id_or_url = parse_uri(uri)

--- a/.github/workflows/bump_deps.yaml
+++ b/.github/workflows/bump_deps.yaml
@@ -56,7 +56,7 @@ jobs:
          rm -rfv ${{ matrix.variable }}_message.txt
          rm -rfv ${{ matrix.variable }}_commit.txt
      - name: Create Pull Request
-        uses: peter-evans/create-pull-request@v7
+        uses: peter-evans/create-pull-request@v6
        with:
          token: ${{ secrets.UPDATE_BOT_TOKEN }}
          push-to-fork: ci-forks/LocalAI
--- a/.github/workflows/bump_docs.yaml
+++ b/.github/workflows/bump_docs.yaml
@@ -17,7 +17,7 @@ jobs:
        run: |
          bash .github/bump_docs.sh ${{ matrix.repository }}
      - name: Create Pull Request
-        uses: peter-evans/create-pull-request@v7
+        uses: peter-evans/create-pull-request@v6
        with:
          token: ${{ secrets.UPDATE_BOT_TOKEN }}
          push-to-fork: ci-forks/LocalAI
--- a/.github/workflows/checksum_checker.yaml
+++ b/.github/workflows/checksum_checker.yaml
@@ -36,7 +36,7 @@ jobs:
          sudo chmod 777 /hf_cache
          bash .github/checksum_checker.sh gallery/index.yaml
      - name: Create Pull Request
-        uses: peter-evans/create-pull-request@v7
+        uses: peter-evans/create-pull-request@v6
        with:
          token: ${{ secrets.UPDATE_BOT_TOKEN }}
          push-to-fork: ci-forks/LocalAI
--- a/.github/workflows/image.yml
+++ b/.github/workflows/image.yml
@@ -13,78 +13,6 @@ concurrency:
  cancel-in-progress: true

 jobs:
-  hipblas-jobs:
-    uses: ./.github/workflows/image_build.yml
-    with:
-      tag-latest: ${{ matrix.tag-latest }}
-      tag-suffix: ${{ matrix.tag-suffix }}
-      ffmpeg: ${{ matrix.ffmpeg }}
-      image-type: ${{ matrix.image-type }}
-      build-type: ${{ matrix.build-type }}
-      cuda-major-version: ${{ matrix.cuda-major-version }}
-      cuda-minor-version: ${{ matrix.cuda-minor-version }}
-      platforms: ${{ matrix.platforms }}
-      runs-on: ${{ matrix.runs-on }}
-      base-image: ${{ matrix.base-image }}
-      grpc-base-image: ${{ matrix.grpc-base-image }}
-      aio: ${{ matrix.aio }}
-      makeflags: ${{ matrix.makeflags }}
-      latest-image: ${{ matrix.latest-image }}
-      latest-image-aio: ${{ matrix.latest-image-aio }}
-    secrets:
-      dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
-      dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
-      quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
-      quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
-    strategy:
-      # Pushing with all jobs in parallel
-      # eats the bandwidth of all the nodes
-      max-parallel: 2
-      matrix:
-        include:
-          - build-type: 'hipblas'
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-hipblas-ffmpeg'
-            ffmpeg: 'true'
-            image-type: 'extras'
-            aio: "-aio-gpu-hipblas"
-            base-image: "rocm/dev-ubuntu-22.04:6.1"
-            grpc-base-image: "ubuntu:22.04"
-            latest-image: 'latest-gpu-hipblas'
-            latest-image-aio: 'latest-aio-gpu-hipblas'
-            runs-on: 'arc-runner-set'
-            makeflags: "--jobs=3 --output-sync=target"
-          - build-type: 'hipblas'
-            platforms: 'linux/amd64'
-            tag-latest: 'false'
-            tag-suffix: '-hipblas'
-            ffmpeg: 'false'
-            image-type: 'extras'
-            base-image: "rocm/dev-ubuntu-22.04:6.1"
-            grpc-base-image: "ubuntu:22.04"
-            runs-on: 'arc-runner-set'
-            makeflags: "--jobs=3 --output-sync=target"
-          - build-type: 'hipblas'
-            platforms: 'linux/amd64'
-            tag-latest: 'false'
-            tag-suffix: '-hipblas-ffmpeg-core'
-            ffmpeg: 'true'
-            image-type: 'core'
-            base-image: "rocm/dev-ubuntu-22.04:6.1"
-            grpc-base-image: "ubuntu:22.04"
-            runs-on: 'arc-runner-set'
-            makeflags: "--jobs=3 --output-sync=target"
-          - build-type: 'hipblas'
-            platforms: 'linux/amd64'
-            tag-latest: 'false'
-            tag-suffix: '-hipblas-core'
-            ffmpeg: 'false'
-            image-type: 'core'
-            base-image: "rocm/dev-ubuntu-22.04:6.1"
-            grpc-base-image: "ubuntu:22.04"
-            runs-on: 'arc-runner-set'
-            makeflags: "--jobs=3 --output-sync=target"
  self-hosted-jobs:
    uses: ./.github/workflows/image_build.yml
    with:
@@ -111,7 +39,7 @@ jobs:
    strategy:
      # Pushing with all jobs in parallel
      # eats the bandwidth of all the nodes
-      max-parallel: ${{ github.event_name != 'pull_request' && 5 || 8 }}
+      max-parallel: ${{ github.event_name != 'pull_request' && 6 || 10 }}
      matrix:
        include:
          # Extra images
@@ -194,6 +122,29 @@ jobs:
            base-image: "ubuntu:22.04"
            runs-on: 'arc-runner-set'
            makeflags: "--jobs=3 --output-sync=target"
+          - build-type: 'hipblas'
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-hipblas-ffmpeg'
+            ffmpeg: 'true'
+            image-type: 'extras'
+            aio: "-aio-gpu-hipblas"
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
+            latest-image: 'latest-gpu-hipblas'
+            latest-image-aio: 'latest-aio-gpu-hipblas'
+            runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
+          - build-type: 'hipblas'
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-hipblas'
+            ffmpeg: 'false'
+            image-type: 'extras'
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
+            runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
          - build-type: 'sycl_f16'
            platforms: 'linux/amd64'
            tag-latest: 'auto'
@@ -261,6 +212,26 @@ jobs:
            image-type: 'core'
            runs-on: 'arc-runner-set'
            makeflags: "--jobs=3 --output-sync=target"
+          - build-type: 'hipblas'
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-hipblas-ffmpeg-core'
+            ffmpeg: 'true'
+            image-type: 'core'
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
+            runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
+          - build-type: 'hipblas'
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-hipblas-core'
+            ffmpeg: 'false'
+            image-type: 'core'
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
+            runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"

  core-image-build:
    uses: ./.github/workflows/image_build.yml
--- a/.github/workflows/release.yaml
+++ b/.github/workflows/release.yaml
@@ -294,7 +294,7 @@ jobs:
          export C_INCLUDE_PATH=/usr/local/include
          export CPLUS_INCLUDE_PATH=/usr/local/include
          export PATH=$PATH:$GOPATH/bin
-          export SKIP_GRPC_BACKEND=backend-assets/grpc/whisper
+
          make dist
      - uses: actions/upload-artifact@v4
        with:
@@ -327,7 +327,7 @@ jobs:
          cache: false
      - name: Dependencies
        run: |
-          brew install protobuf grpc libomp llvm
+          brew install protobuf grpc
          go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af
          go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.2
      - name: Build
@@ -336,7 +336,7 @@ jobs:
          export C_INCLUDE_PATH=/usr/local/include
          export CPLUS_INCLUDE_PATH=/usr/local/include
          export PATH=$PATH:$GOPATH/bin
-          export CC=/opt/homebrew/opt/llvm/bin/clang
+
          make dist
      - uses: actions/upload-artifact@v4
        with:
--- a/.github/workflows/secscan.yaml
+++ b/.github/workflows/secscan.yaml
@@ -18,7 +18,7 @@ jobs:
        if: ${{ github.actor != 'dependabot[bot]' }}
      - name: Run Gosec Security Scanner
        if: ${{ github.actor != 'dependabot[bot]' }}
-        uses: securego/gosec@v2.21.0
+        uses: securego/gosec@master
        with:
          # we let the report trigger content trigger a failure using the GitHub Security features.
          args: '-no-fail -fmt sarif -out results.sarif ./...'
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -214,13 +214,12 @@ jobs:
        run: go version
      - name: Dependencies
        run: |
-          brew install protobuf grpc make protoc-gen-go protoc-gen-go-grpc libomp llvm
+          brew install protobuf grpc make protoc-gen-go protoc-gen-go-grpc
          pip install --user --no-cache-dir grpcio-tools==1.64.1
      - name: Test
        run: |
          export C_INCLUDE_PATH=/usr/local/include
          export CPLUS_INCLUDE_PATH=/usr/local/include
-          export CC=/opt/homebrew/opt/llvm/bin/clang
          # Used to run the newer GNUMake version from brew that supports --output-sync
          export PATH="/opt/homebrew/opt/make/libexec/gnubin:$PATH"
          BUILD_TYPE="GITHUB_CI_HAS_BROKEN_METAL" CMAKE_ARGS="-DGGML_F16C=OFF -DGGML_AVX512=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF" make --jobs 4 --output-sync=target test
--- a/.github/workflows/update_swagger.yaml
+++ b/.github/workflows/update_swagger.yaml
@@ -25,7 +25,7 @@ jobs:
        run: |
          make protogen-go swagger
      - name: Create Pull Request
-        uses: peter-evans/create-pull-request@v7
+        uses: peter-evans/create-pull-request@v6
        with:
          token: ${{ secrets.UPDATE_BOT_TOKEN }}
          push-to-fork: ci-forks/LocalAI
--- a/44
+++ b/44
@@ -13,7 +13,7 @@ ARG TARGETARCH
 ARG TARGETVARIANT

 ENV DEBIAN_FRONTEND=noninteractive
-ENV EXTERNAL_GRPC_BACKENDS="coqui:/build/backend/python/coqui/run.sh,huggingface-embeddings:/build/backend/python/sentencetransformers/run.sh,transformers:/build/backend/python/transformers/run.sh,sentencetransformers:/build/backend/python/sentencetransformers/run.sh,rerankers:/build/backend/python/rerankers/run.sh,autogptq:/build/backend/python/autogptq/run.sh,bark:/build/backend/python/bark/run.sh,diffusers:/build/backend/python/diffusers/run.sh,openvoice:/build/backend/python/openvoice/run.sh,vall-e-x:/build/backend/python/vall-e-x/run.sh,vllm:/build/backend/python/vllm/run.sh,mamba:/build/backend/python/mamba/run.sh,exllama2:/build/backend/python/exllama2/run.sh,transformers-musicgen:/build/backend/python/transformers-musicgen/run.sh,parler-tts:/build/backend/python/parler-tts/run.sh"
+ENV EXTERNAL_GRPC_BACKENDS="coqui:/build/backend/python/coqui/run.sh,huggingface-embeddings:/build/backend/python/sentencetransformers/run.sh,transformers:/build/backend/python/transformers/run.sh,sentencetransformers:/build/backend/python/sentencetransformers/run.sh,rerankers:/build/backend/python/rerankers/run.sh,autogptq:/build/backend/python/autogptq/run.sh,bark:/build/backend/python/bark/run.sh,diffusers:/build/backend/python/diffusers/run.sh,exllama:/build/backend/python/exllama/run.sh,openvoice:/build/backend/python/openvoice/run.sh,vall-e-x:/build/backend/python/vall-e-x/run.sh,vllm:/build/backend/python/vllm/run.sh,mamba:/build/backend/python/mamba/run.sh,exllama2:/build/backend/python/exllama2/run.sh,transformers-musicgen:/build/backend/python/transformers-musicgen/run.sh,parler-tts:/build/backend/python/parler-tts/run.sh"


 RUN apt-get update && \
@@ -263,20 +263,14 @@ EOT
 # In most cases, builder is the image you should be using - however, this can save build time if one just needs to copy backend-assets/grpc/stablediffusion and nothing else.
 FROM builder-base AS builder-sd

-# stablediffusion does not tolerate a newer version of abseil, copy only over enough elements to build it
-COPY Makefile .
-COPY go.mod .
-COPY go.sum .
-COPY backend/backend.proto ./backend/backend.proto
-COPY backend/go/image/stablediffusion ./backend/go/image/stablediffusion
-COPY pkg/grpc ./pkg/grpc
-COPY pkg/stablediffusion ./pkg/stablediffusion
-RUN git init
-RUN make sources/go-stable-diffusion
-RUN touch prepare-sources
+COPY . .
+COPY .git .

-# Actually build the backend
-RUN GRPC_BACKENDS=backend-assets/grpc/stablediffusion make backend-assets/grpc/stablediffusion
+RUN make prepare
+
+
+# stablediffusion does not tolerate a newer version of abseil, build it first
+RUN GRPC_BACKENDS=backend-assets/grpc/stablediffusion make build

 ###################################
 ###################################
@@ -291,20 +285,8 @@ COPY --from=grpc /opt/grpc /usr/local
 # Rebuild with defaults backends
 WORKDIR /build

-COPY . .
-COPY .git .
-
-RUN make prepare
-
 ## Build the binary
-## If it's CUDA or hipblas, we want to skip some of the llama-compat backends to save space
-## We only leave the most CPU-optimized variant and the fallback for the cublas/hipblas build
-## (both will use CUDA or hipblas for the actual computation)
-RUN if [ "${BUILD_TYPE}" = "cublas" ] || [ "${BUILD_TYPE}" = "hipblas" ]; then \
-        SKIP_GRPC_BACKEND="backend-assets/grpc/llama-cpp-avx backend-assets/grpc/llama-cpp-avx2" make build; \
-    else \
-        make build; \
-    fi
+RUN make build

 RUN if [ ! -d "/build/sources/go-piper/piper-phonemize/pi/lib/" ]; then \
        mkdir -p /build/sources/go-piper/piper-phonemize/pi/lib/ \
@@ -338,8 +320,9 @@ RUN if [ "${FFMPEG}" = "true" ]; then \

 RUN apt-get update && \
    apt-get install -y --no-install-recommends \
-        ssh less wget
-# For the devcontainer, leave apt functional in case additional devtools are needed at runtime.
+        ssh less && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*

 RUN go install github.com/go-delve/delve/cmd/dlv@latest

@@ -417,6 +400,9 @@ RUN if [[ ( "${EXTRA_BACKENDS}" =~ "coqui" || -z "${EXTRA_BACKENDS}" ) && "$IMAG
    ; fi && \
    if [[ ( "${EXTRA_BACKENDS}" =~ "transformers-musicgen" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
        make -C backend/python/transformers-musicgen \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "exllama1" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/exllama \
    ; fi

 RUN if [[ ( "${EXTRA_BACKENDS}" =~ "vall-e-x" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
--- a/24
+++ b/24
@@ -8,7 +8,7 @@ DETECT_LIBS?=true
 # llama.cpp versions
 GOLLAMA_REPO?=https://github.com/go-skynet/go-llama.cpp
 GOLLAMA_VERSION?=2b57a8ae43e4699d3dc5d1496a1ccd42922993be
-CPPLLAMA_VERSION?=f0c7b5edf82aa200656fd88c11ae3a805d7130bf
+CPPLLAMA_VERSION?=2f3c1466ff46a2413b0e363a5005c46538186ee6

 # go-rwkv version
 RWKV_REPO?=https://github.com/donomii/go-rwkv.cpp
@@ -16,7 +16,7 @@ RWKV_VERSION?=661e7ae26d442f5cfebd2a0881b44e8c55949ec6

 # whisper.cpp version
 WHISPER_REPO?=https://github.com/ggerganov/whisper.cpp
-WHISPER_CPP_VERSION?=34972dbe221709323714fc8402f2e24041d48213
+WHISPER_CPP_VERSION?=d65786ea540a5aef21f67cacfa6f134097727780

 # bert.cpp version
 BERT_REPO?=https://github.com/go-skynet/go-bert.cpp
@@ -338,7 +338,7 @@ rebuild: ## Rebuilds the project
 	$(MAKE) -C sources/go-tiny-dream clean
 	$(MAKE) build

-prepare: prepare-sources $(OPTIONAL_TARGETS)
+prepare: prepare-sources gen-assets $(OPTIONAL_TARGETS)

 clean: ## Remove build related file
 	$(GOCMD) clean -cache
@@ -359,9 +359,6 @@ clean-tests:
 	rm -rf test-dir
 	rm -rf core/http/backend-assets

-clean-dc: clean
-	cp -r /build/backend-assets /workspace/backend-assets
-
 ## Build:
 build: prepare backend-assets grpcs ## Build the project
 	$(info ${GREEN}I local-ai build info:${RESET})
@@ -537,10 +534,10 @@ protogen-go-clean:
 	$(RM) bin/*

 .PHONY: protogen-python
-protogen-python: autogptq-protogen bark-protogen coqui-protogen diffusers-protogen exllama2-protogen mamba-protogen rerankers-protogen sentencetransformers-protogen transformers-protogen parler-tts-protogen transformers-musicgen-protogen vall-e-x-protogen vllm-protogen openvoice-protogen
+protogen-python: autogptq-protogen bark-protogen coqui-protogen diffusers-protogen exllama-protogen exllama2-protogen mamba-protogen rerankers-protogen sentencetransformers-protogen transformers-protogen parler-tts-protogen transformers-musicgen-protogen vall-e-x-protogen vllm-protogen openvoice-protogen

 .PHONY: protogen-python-clean
-protogen-python-clean: autogptq-protogen-clean bark-protogen-clean coqui-protogen-clean diffusers-protogen-clean  exllama2-protogen-clean mamba-protogen-clean sentencetransformers-protogen-clean rerankers-protogen-clean transformers-protogen-clean transformers-musicgen-protogen-clean parler-tts-protogen-clean vall-e-x-protogen-clean vllm-protogen-clean openvoice-protogen-clean
+protogen-python-clean: autogptq-protogen-clean bark-protogen-clean coqui-protogen-clean diffusers-protogen-clean exllama-protogen-clean exllama2-protogen-clean mamba-protogen-clean sentencetransformers-protogen-clean rerankers-protogen-clean transformers-protogen-clean transformers-musicgen-protogen-clean parler-tts-protogen-clean vall-e-x-protogen-clean vllm-protogen-clean openvoice-protogen-clean

 .PHONY: autogptq-protogen
 autogptq-protogen:
@@ -574,6 +571,14 @@ diffusers-protogen:
 diffusers-protogen-clean:
 	$(MAKE) -C backend/python/diffusers protogen-clean

+.PHONY: exllama-protogen
+exllama-protogen:
+	$(MAKE) -C backend/python/exllama protogen
+
+.PHONY: exllama-protogen-clean
+exllama-protogen-clean:
+	$(MAKE) -C backend/python/exllama protogen-clean
+
 .PHONY: exllama2-protogen
 exllama2-protogen:
 	$(MAKE) -C backend/python/exllama2 protogen
@@ -670,6 +675,7 @@ prepare-extra-conda-environments: protogen-python
 	$(MAKE) -C backend/python/parler-tts
 	$(MAKE) -C backend/python/vall-e-x
 	$(MAKE) -C backend/python/openvoice
+	$(MAKE) -C backend/python/exllama
 	$(MAKE) -C backend/python/exllama2

 prepare-test-extra: protogen-python
@@ -840,7 +846,7 @@ endif

 backend-assets/grpc/whisper: sources/whisper.cpp sources/whisper.cpp/libwhisper.a backend-assets/grpc
 	CGO_LDFLAGS="$(CGO_LDFLAGS) $(CGO_LDFLAGS_WHISPER)" C_INCLUDE_PATH="$(CURDIR)/sources/whisper.cpp/include:$(CURDIR)/sources/whisper.cpp/ggml/include" LIBRARY_PATH=$(CURDIR)/sources/whisper.cpp \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/whisper ./backend/go/transcribe/whisper
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/whisper ./backend/go/transcribe/
 ifneq ($(UPX),)
 	$(UPX) backend-assets/grpc/whisper
 endif
--- a/README.md
+++ b/README.md
@@ -40,7 +40,7 @@

 > :bulb: Get help - [❓FAQ](https://localai.io/faq/) [💭Discussions](https://github.com/go-skynet/LocalAI/discussions) [:speech_balloon: Discord](https://discord.gg/uJAeKSAGDy) [:book: Documentation website](https://localai.io/)
 >
-> [💻 Quickstart](https://localai.io/basics/getting_started/) [🖼️ Models](https://models.localai.io/) [🚀 Roadmap](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap) [🥽 Demo](https://demo.localai.io) [🌍 Explorer](https://explorer.localai.io) [🛫 Examples](https://github.com/go-skynet/LocalAI/tree/master/examples/) 
+> [💻 Quickstart](https://localai.io/basics/getting_started/) [📣 News](https://localai.io/basics/news/) [ 🛫 Examples ](https://github.com/go-skynet/LocalAI/tree/master/examples/) [ 🖼️ Models ](https://localai.io/models/) [ 🚀 Roadmap ](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap)

 [![tests](https://github.com/go-skynet/LocalAI/actions/workflows/test.yml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/test.yml)[![Build and Release](https://github.com/go-skynet/LocalAI/actions/workflows/release.yaml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/release.yaml)[![build container images](https://github.com/go-skynet/LocalAI/actions/workflows/image.yml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/image.yml)[![Bump dependencies](https://github.com/go-skynet/LocalAI/actions/workflows/bump_deps.yaml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/bump_deps.yaml)[![Artifact Hub](https://img.shields.io/endpoint?url=https://artifacthub.io/badge/repository/localai)](https://artifacthub.io/packages/search?repo=localai)

@@ -72,7 +72,6 @@ docker run -ti --name local-ai -p 8080:8080 localai/localai:latest-aio-cpu

 [Roadmap](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap)

- Aug 2024:  🆕 FLUX-1, [P2P Explorer](https://explorer.localai.io)
 - July 2024: 🔥🔥 🆕 P2P Dashboard, LocalAI Federated mode and AI Swarms: https://github.com/mudler/LocalAI/pull/2723
 - June 2024: 🆕 You can browse now the model gallery without LocalAI! Check out https://models.localai.io
 - June 2024: Support for models from OCI registries: https://github.com/mudler/LocalAI/pull/2628
--- a/aio/cpu/vision.yaml
+++ b/aio/cpu/vision.yaml
@@ -2,7 +2,7 @@ backend: llama-cpp
 context_size: 4096
 f16: true
 mmap: true
-name: gpt-4o
+name: gpt-4-vision-preview

 roles:
  user: "USER:"
--- a/aio/gpu-8g/vision.yaml
+++ b/aio/gpu-8g/vision.yaml
@@ -2,7 +2,7 @@ backend: llama-cpp
 context_size: 4096
 f16: true
 mmap: true
-name: gpt-4o
+name: gpt-4-vision-preview

 roles:
  user: "USER:"
--- a/aio/intel/image-gen.yaml
+++ b/aio/intel/image-gen.yaml
@@ -1,6 +1,6 @@
 name: stablediffusion
 parameters:
-  model: Lykon/dreamshaper-8
+  model: runwayml/stable-diffusion-v1-5
 backend: diffusers
 step: 25
 f16: true
--- a/aio/intel/vision.yaml
+++ b/aio/intel/vision.yaml
@@ -2,7 +2,7 @@ backend: llama-cpp
 context_size: 4096
 mmap: false
 f16: false
-name: gpt-4o
+name: gpt-4-vision-preview

 roles:
  user: "USER:"
--- a/backend/backend.proto
+++ b/backend/backend.proto
@@ -16,7 +16,6 @@ service Backend {
  rpc GenerateImage(GenerateImageRequest) returns (Result) {}
  rpc AudioTranscription(TranscriptRequest) returns (TranscriptResult) {}
  rpc TTS(TTSRequest) returns (Result) {}
-  rpc SoundGeneration(SoundGenerationRequest) returns (Result) {}
  rpc TokenizeString(PredictOptions) returns (TokenizationResponse) {}
  rpc Status(HealthMessage) returns (StatusResponse) {}

@@ -134,8 +133,6 @@ message PredictOptions {
  repeated string Images = 42;
  bool UseTokenizerTemplate = 43;
  repeated Message Messages = 44;
-  repeated string Videos = 45;
-  repeated string Audios = 46;
 }

 // The response message containing the result
@@ -273,17 +270,6 @@ message TTSRequest {
  optional string language = 5;
 }

-message SoundGenerationRequest {
-  string text = 1;
-  string model = 2;
-  string dst = 3;
-  optional float duration = 4;
-  optional float temperature = 5;
-  optional bool sample = 6;
-  optional string src = 7;
-  optional int32 src_divisor = 8;
-}
-
 message TokenizationResponse {
  int32 length = 1;
  repeated int32 tokens = 2;
--- a/backend/cpp/llama/grpc-server.cpp
+++ b/backend/cpp/llama/grpc-server.cpp
@@ -13,15 +13,15 @@
 #include <getopt.h>
 #include "clip.h"
 #include "llava.h"
-#include "log.h"
 #include "stb_image.h"
 #include "common.h"
 #include "json.hpp"
 #include "llama.h"
+#include "grammar-parser.h"
 #include "backend.pb.h"
 #include "backend.grpc.pb.h"
 #include "utils.hpp"
-#include "sampling.h"
+
 // include std::regex
 #include <cstddef>
 #include <thread>
@@ -203,8 +203,8 @@ struct llama_client_slot
    std::string stopping_word;

    // sampling
-    struct gpt_sampler_params sparams;
-    gpt_sampler *ctx_sampling = nullptr;
+    struct llama_sampling_params sparams;
+    llama_sampling_context *ctx_sampling = nullptr;

    int32_t ga_i = 0;   // group-attention state
    int32_t ga_n = 1;   // group-attention factor
@@ -449,7 +449,7 @@ struct llama_server_context
            LOG_INFO("Multi Modal Mode Enabled", {});
            clp_ctx = clip_model_load(params.mmproj.c_str(), /*verbosity=*/ 1);
            if(clp_ctx == nullptr) {
-                LOG_ERR("unable to load clip model: %s", params.mmproj.c_str());
+                LOG_ERROR("unable to load clip model", {{"model", params.mmproj}});
                return false;
            }

@@ -463,7 +463,7 @@ struct llama_server_context
        ctx = llama_init.context;
        if (model == nullptr)
        {
-            LOG_ERR("unable to load model: %s", params.model.c_str());
+            LOG_ERROR("unable to load model", {{"model", params.model}});
            return false;
        }

@@ -471,7 +471,7 @@ struct llama_server_context
            const int n_embd_clip = clip_n_mmproj_embd(clp_ctx);
            const int n_embd_llm  = llama_n_embd(model);
            if (n_embd_clip != n_embd_llm) {
-                LOG("%s: embedding dim of the multimodal projector (%d) is not equal to that of LLaMA (%d). Make sure that you use the correct mmproj file.\n", __func__, n_embd_clip, n_embd_llm);
+                LOG_TEE("%s: embedding dim of the multimodal projector (%d) is not equal to that of LLaMA (%d). Make sure that you use the correct mmproj file.\n", __func__, n_embd_clip, n_embd_llm);
                llama_free(ctx);
                llama_free_model(model);
                return false;
@@ -490,7 +490,7 @@ struct llama_server_context
        std::vector<char> buf(1);
        int res = llama_chat_apply_template(model, nullptr, chat, 1, true, buf.data(), buf.size());
        if (res < 0) {
-            LOG_ERR("The chat template comes with this model is not yet supported, falling back to chatml. This may cause the model to output suboptimal responses", __func__);
+            LOG_ERROR("The chat template comes with this model is not yet supported, falling back to chatml. This may cause the model to output suboptimal responses", {});
            sparams.chat_template = "<|im_start|>"; // llama_chat_apply_template only checks if <|im_start|> exist in the template
        }
    }
@@ -619,7 +619,7 @@ struct llama_server_context

    bool launch_slot_with_data(llama_client_slot* &slot, json data) {
        slot_params default_params;
-        gpt_sampler_params default_sparams;
+        llama_sampling_params default_sparams;
 
        slot->params.stream             = json_value(data, "stream",            false);
        slot->params.cache_prompt       = json_value(data, "cache_prompt",      false);
@@ -628,7 +628,7 @@ struct llama_server_context
        slot->sparams.top_p             = json_value(data, "top_p",             default_sparams.top_p);
        slot->sparams.min_p             = json_value(data, "min_p",             default_sparams.min_p);
        slot->sparams.tfs_z             = json_value(data, "tfs_z",             default_sparams.tfs_z);
-        slot->sparams.typ_p             = json_value(data, "typical_p",         default_sparams.typ_p);
+        slot->sparams.typical_p         = json_value(data, "typical_p",         default_sparams.typical_p);
        slot->sparams.temp              = json_value(data, "temperature",       default_sparams.temp);
        slot->sparams.dynatemp_range    = json_value(data, "dynatemp_range",    default_sparams.dynatemp_range);
        slot->sparams.dynatemp_exponent = json_value(data, "dynatemp_exponent", default_sparams.dynatemp_exponent);
@@ -641,7 +641,7 @@ struct llama_server_context
        slot->sparams.mirostat_eta      = json_value(data, "mirostat_eta",      default_sparams.mirostat_eta);
        slot->sparams.penalize_nl       = json_value(data, "penalize_nl",       default_sparams.penalize_nl);
        slot->params.n_keep             = json_value(data, "n_keep",            slot->params.n_keep);
-        slot->sparams.seed               = json_value(data, "seed",              default_sparams.seed);
+        slot->params.seed               = json_value(data, "seed",              default_params.seed);
        slot->sparams.grammar           = json_value(data, "grammar",           default_sparams.grammar);
        slot->sparams.n_probs           = json_value(data, "n_probs",           default_sparams.n_probs);
        slot->sparams.min_keep          = json_value(data, "min_keep",          default_sparams.min_keep);
@@ -665,7 +665,6 @@ struct llama_server_context
            slot->params.input_prefix = "";
        }

-
        if (data.count("input_suffix") != 0)
        {
            slot->params.input_suffix = data["input_suffix"];
@@ -684,10 +683,6 @@ struct llama_server_context
            slot->prompt = "";
        }

-        if (json_value(data, "ignore_eos", false)) {
-                slot->sparams.logit_bias.push_back({llama_token_eos(model), -INFINITY});
-        }
-        /*
        slot->sparams.penalty_prompt_tokens.clear();
        slot->sparams.use_penalty_prompt_tokens = false;
        const auto &penalty_prompt = data.find("penalty_prompt");
@@ -723,10 +718,14 @@ struct llama_server_context
                slot->sparams.use_penalty_prompt_tokens = true;
            }
        }
-      */

        slot->sparams.logit_bias.clear();

+        if (json_value(data, "ignore_eos", false))
+        {
+            slot->sparams.logit_bias[llama_token_eos(model)] = -INFINITY;
+        }
+
        const auto &logit_bias = data.find("logit_bias");
        if (logit_bias != data.end() && logit_bias->is_array())
        {
@@ -754,7 +753,7 @@ struct llama_server_context
                        llama_token tok = el[0].get<llama_token>();
                        if (tok >= 0 && tok < n_vocab)
                        {
-                            slot->sparams.logit_bias.push_back({tok, bias});
+                            slot->sparams.logit_bias[tok] = bias;
                        }
                    }
                    else if (el[0].is_string())
@@ -762,13 +761,13 @@ struct llama_server_context
                        auto toks = llama_tokenize(model, el[0].get<std::string>(), false);
                        for (auto tok : toks)
                        {
-                            slot->sparams.logit_bias.push_back({tok, bias});
+                            slot->sparams.logit_bias[tok] = bias;
                        }
                    }
                }
            }
        }
-        
+
        slot->params.antiprompt.clear();

        const auto &stop = data.find("stop");
@@ -782,22 +781,24 @@ struct llama_server_context
                }
            }
        }
-        
-        const auto & samplers = data.find("samplers");
-        if (samplers != data.end() && samplers->is_array()) {
+
+        const auto &samplers_sequence = data.find("samplers");
+        if (samplers_sequence != data.end() && samplers_sequence->is_array())
+        {
            std::vector<std::string> sampler_names;
-                for (const auto & name : *samplers) {
-                    if (name.is_string()) {
-                        sampler_names.emplace_back(name);
-                    }
+            for (const auto &sampler_name : *samplers_sequence)
+            {
+                if (sampler_name.is_string())
+                {
+                    sampler_names.emplace_back(sampler_name);
                }
-                slot->sparams.samplers = gpt_sampler_types_from_names(sampler_names, false);
+            }
+            slot->sparams.samplers_sequence = llama_sampling_types_from_names(sampler_names, false);
        }
        else
        {
-                slot->sparams.samplers = default_sparams.samplers;
+            slot->sparams.samplers_sequence = default_sparams.samplers_sequence;
        }
-        

        if (multimodal)
        {
@@ -813,11 +814,10 @@ struct llama_server_context
                    img_sl.img_data = clip_image_u8_init();
                    if (!clip_image_load_from_bytes(image_buffer.data(), image_buffer.size(), img_sl.img_data))
                    {
-                        LOG_ERR("%s: failed to load image, slot_id: %d, img_sl_id: %d", 
-                             __func__,
-                             slot->id,
-                             img_sl.id
-                        );
+                        LOG_ERROR("failed to load image", {
+                            {"slot_id",   slot->id},
+                            {"img_sl_id", img_sl.id}
+                        });
                        return false;
                    }
                    LOG_VERBOSE("image loaded", {
@@ -855,12 +855,12 @@ struct llama_server_context
                                    }
                                }
                                if (!found) {
-                                    LOG("ERROR: Image with id: %i, not found.\n", img_id);
+                                    LOG_TEE("ERROR: Image with id: %i, not found.\n", img_id);
                                    slot->images.clear();
                                    return false;
                                }
                            } catch (const std::invalid_argument& e) {
-                                LOG("Invalid image number id in prompt\n");
+                                LOG_TEE("Invalid image number id in prompt\n");
                                slot->images.clear();
                                return false;
                            }
@@ -875,10 +875,10 @@ struct llama_server_context

        if (slot->ctx_sampling != nullptr)
        {
-            gpt_sampler_free(slot->ctx_sampling);
+            llama_sampling_free(slot->ctx_sampling);
        }
-        slot->ctx_sampling = gpt_sampler_init(model, slot->sparams);
-        //llama_set_rng_seed(ctx, slot->params.seed);
+        slot->ctx_sampling = llama_sampling_init(slot->sparams);
+        llama_set_rng_seed(ctx, slot->params.seed);
        slot->command = LOAD_PROMPT;

        all_slots_are_idle = false;
@@ -888,7 +888,7 @@ struct llama_server_context
            {"task_id", slot->task_id},
        });

-      //  LOG("sampling: \n%s\n", llama_sampling_print(slot->sparams).c_str());
+        LOG_TEE("sampling: \n%s\n", llama_sampling_print(slot->sparams).c_str());

        return true;
    }
@@ -928,7 +928,7 @@ struct llama_server_context
                };
                if (llama_decode(ctx, batch_view) != 0)
                {
-                    LOG("%s: llama_decode() failed\n", __func__);
+                    LOG_TEE("%s: llama_decode() failed\n", __func__);
                    return;
                }
            }
@@ -940,7 +940,7 @@ struct llama_server_context
            }
        }

-        LOG("system prompt updated\n");
+        LOG_TEE("system prompt updated\n");
        system_need_update = false;
    }

@@ -1006,13 +1006,11 @@ struct llama_server_context
        slot.generated_text += token_str;
        slot.has_next_token = true;

-/*
        if (slot.ctx_sampling->params.use_penalty_prompt_tokens && result.tok != -1)
        {
            // we can change penalty_prompt_tokens because it is always created from scratch each request
            slot.ctx_sampling->params.penalty_prompt_tokens.push_back(result.tok);
        }
-        */

        // check if there is incomplete UTF-8 character at the end
        bool incomplete = false;
@@ -1121,8 +1119,8 @@ struct llama_server_context
                continue;
            }

-            if (!llava_image_embed_make_with_clip_img(clp_ctx, params.cpuparams.n_threads, img.img_data, &img.image_embedding, &img.image_tokens)) {
-                LOG("Error processing the given image");
+            if (!llava_image_embed_make_with_clip_img(clp_ctx, params.n_threads, img.img_data, &img.image_embedding, &img.image_tokens)) {
+                LOG_TEE("Error processing the given image");
                return false;
            }

@@ -1134,7 +1132,7 @@ struct llama_server_context

    void send_error(task_server& task, const std::string &error)
    {
-        LOG("task %i - error: %s\n", task.id, error.c_str());
+        LOG_TEE("task %i - error: %s\n", task.id, error.c_str());
        task_result res;
        res.id = task.id;
        res.multitask_id = task.multitask_id;
@@ -1146,11 +1144,13 @@ struct llama_server_context

    json get_formated_generation(llama_client_slot &slot)
    {
-        std::vector<std::string> samplers;
-        samplers.reserve(slot.sparams.samplers.size());
-        for (const auto & sampler : slot.sparams.samplers)
+        const auto eos_bias = slot.sparams.logit_bias.find(llama_token_eos(model));
+        const bool ignore_eos = eos_bias != slot.sparams.logit_bias.end() &&
+                                eos_bias->second < 0.0f && std::isinf(eos_bias->second);
+        std::vector<std::string> samplers_sequence;
+        for (const auto &sampler_type : slot.sparams.samplers_sequence)
        {
-            samplers.emplace_back(gpt_sampler_type_to_str(sampler));
+            samplers_sequence.emplace_back(llama_sampling_type_to_str(sampler_type));
        }

        return json {
@@ -1165,11 +1165,13 @@ struct llama_server_context
            {"top_p",             slot.sparams.top_p},
            {"min_p",             slot.sparams.min_p},
            {"tfs_z",             slot.sparams.tfs_z},
-            {"typical_p",         slot.sparams.typ_p},
+            {"typical_p",         slot.sparams.typical_p},
            {"repeat_last_n",     slot.sparams.penalty_last_n},
            {"repeat_penalty",    slot.sparams.penalty_repeat},
            {"presence_penalty",  slot.sparams.penalty_present},
            {"frequency_penalty", slot.sparams.penalty_freq},
+            {"penalty_prompt_tokens", slot.sparams.penalty_prompt_tokens},
+            {"use_penalty_prompt_tokens", slot.sparams.use_penalty_prompt_tokens},
            {"mirostat",          slot.sparams.mirostat},
            {"mirostat_tau",      slot.sparams.mirostat_tau},
            {"mirostat_eta",      slot.sparams.mirostat_eta},
@@ -1177,13 +1179,13 @@ struct llama_server_context
            {"stop",              slot.params.antiprompt},
            {"n_predict",         slot.params.n_predict},
            {"n_keep",            params.n_keep},
-            {"ignore_eos",        slot.sparams.ignore_eos},
+            {"ignore_eos",        ignore_eos},
            {"stream",            slot.params.stream},
-      //      {"logit_bias",        slot.sparams.logit_bias},
+            {"logit_bias",        slot.sparams.logit_bias},
            {"n_probs",           slot.sparams.n_probs},
            {"min_keep",          slot.sparams.min_keep},
            {"grammar",           slot.sparams.grammar},
-            {"samplers",          samplers}
+            {"samplers",          samplers_sequence}
        };
    }

@@ -1373,7 +1375,7 @@ struct llama_server_context
                };
                if (llama_decode(ctx, batch_view))
                {
-                    LOG("%s : failed to eval\n", __func__);
+                    LOG_TEE("%s : failed to eval\n", __func__);
                    return false;
                }
            }
@@ -1391,7 +1393,7 @@ struct llama_server_context
                llama_batch batch_img = { n_eval, nullptr, (img.image_embedding + i * n_embd), nullptr, nullptr, nullptr, nullptr, slot.n_past, 1, 0, };
                if (llama_decode(ctx, batch_img))
                {
-                    LOG("%s : failed to eval image\n", __func__);
+                    LOG_TEE("%s : failed to eval image\n", __func__);
                    return false;
                }
                slot.n_past += n_eval;
@@ -1574,7 +1576,7 @@ struct llama_server_context
                    slot.n_past = 0;
                    slot.truncated = false;
                    slot.has_next_token = true;
-                    LOG("Context exhausted. Slot %d released (%d tokens in cache)\n", slot.id, (int) slot.cache_tokens.size());
+                    LOG_TEE("Context exhausted. Slot %d released (%d tokens in cache)\n", slot.id, (int) slot.cache_tokens.size());

                    continue;
                    // END LOCALAI changes
@@ -1712,7 +1714,7 @@ struct llama_server_context

                    if (!slot.params.cache_prompt)
                    {
-                        gpt_sampler_reset(slot.ctx_sampling);
+                        llama_sampling_reset(slot.ctx_sampling);

                        slot.n_past = 0;
                        slot.n_past_se = 0;
@@ -1724,7 +1726,7 @@ struct llama_server_context
                        // push the prompt into the sampling context (do not apply grammar)
                        for (auto &token : prompt_tokens)
                        {
-                            gpt_sampler_accept(slot.ctx_sampling, token, false);
+                            llama_sampling_accept(slot.ctx_sampling, ctx, token, false);
                        }

                        slot.n_past = common_part(slot.cache_tokens, prompt_tokens);
@@ -1822,11 +1824,10 @@ struct llama_server_context

                    if (has_images && !ingest_images(slot, n_batch))
                    {
-                        LOG_ERR("%s: failed processing images Slot id : %d, Task id: %d", 
-                            __func__,
-                            slot.id,
-                            slot.task_id
-                        );
+                        LOG_ERROR("failed processing images", {
+                            "slot_id", slot.id,
+                            "task_id", slot.task_id,
+                        });
                        // FIXME @phymbert: to be properly tested
                        //  early returning without changing the slot state will block the slot for ever
                        // no one at the moment is checking the return value
@@ -1866,10 +1867,10 @@ struct llama_server_context
                        const int bd = (slot.ga_w / slot.ga_n) * (slot.ga_n - 1);
                        const int dd = (slot.ga_w / slot.ga_n) - ib * bd - slot.ga_w;

-                        LOG("\n");
-                        LOG("shift: [%6d, %6d] + %6d -> [%6d, %6d]\n", slot.ga_i, slot.n_past_se, ib * bd, slot.ga_i + ib * bd, slot.n_past_se + ib * bd);
-                        LOG("div:   [%6d, %6d] / %6d -> [%6d, %6d]\n", slot.ga_i + ib * bd, slot.ga_i + ib * bd + slot.ga_w, slot.ga_n, (slot.ga_i + ib * bd) / slot.ga_n, (slot.ga_i + ib * bd + slot.ga_w) / slot.ga_n);
-                        LOG("shift: [%6d, %6d] + %6d -> [%6d, %6d]\n", slot.ga_i + ib * bd + slot.ga_w, slot.n_past_se + ib * bd, dd, slot.ga_i + ib * bd + slot.ga_w + dd, slot.n_past_se + ib * bd + dd);
+                        LOG_TEE("\n");
+                        LOG_TEE("shift: [%6d, %6d] + %6d -> [%6d, %6d]\n", slot.ga_i, slot.n_past_se, ib * bd, slot.ga_i + ib * bd, slot.n_past_se + ib * bd);
+                        LOG_TEE("div:   [%6d, %6d] / %6d -> [%6d, %6d]\n", slot.ga_i + ib * bd, slot.ga_i + ib * bd + slot.ga_w, slot.ga_n, (slot.ga_i + ib * bd) / slot.ga_n, (slot.ga_i + ib * bd + slot.ga_w) / slot.ga_n);
+                        LOG_TEE("shift: [%6d, %6d] + %6d -> [%6d, %6d]\n", slot.ga_i + ib * bd + slot.ga_w, slot.n_past_se + ib * bd, dd, slot.ga_i + ib * bd + slot.ga_w + dd, slot.n_past_se + ib * bd + dd);

                        llama_kv_cache_seq_add(ctx, slot.id, slot.ga_i, slot.n_past_se, ib * bd);
                        llama_kv_cache_seq_div(ctx, slot.id, slot.ga_i + ib * bd, slot.ga_i + ib * bd + slot.ga_w,slot.ga_n);
@@ -1879,7 +1880,7 @@ struct llama_server_context

                        slot.ga_i += slot.ga_w / slot.ga_n;

-                        LOG("\nn_past_old = %d, n_past = %d, ga_i = %d\n\n", slot.n_past_se + bd, slot.n_past_se, slot.ga_i);
+                        LOG_TEE("\nn_past_old = %d, n_past = %d, ga_i = %d\n\n", slot.n_past_se + bd, slot.n_past_se, slot.ga_i);
                    }
                    slot.n_past_se += n_tokens;
                }
@@ -1904,11 +1905,11 @@ struct llama_server_context
                if (n_batch == 1 || ret < 0)
                {
                    // if you get here, it means the KV cache is full - try increasing it via the context size
-                    LOG("%s : failed to decode the batch, n_batch = %d, ret = %d\n", __func__, n_batch, ret);
+                    LOG_TEE("%s : failed to decode the batch, n_batch = %d, ret = %d\n", __func__, n_batch, ret);
                    return false;
                }

-                LOG("%s : failed to find free space in the KV cache, retrying with smaller n_batch = %d\n", __func__, n_batch / 2);
+                LOG_TEE("%s : failed to find free space in the KV cache, retrying with smaller n_batch = %d\n", __func__, n_batch / 2);

                // retry with half the batch size to try to find a free slot in the KV cache
                n_batch /= 2;
@@ -1933,9 +1934,9 @@ struct llama_server_context
                }

                completion_token_output result;
-                const llama_token id = gpt_sampler_sample(slot.ctx_sampling, ctx, slot.i_batch - i);
+                const llama_token id = llama_sampling_sample(slot.ctx_sampling, ctx, NULL, slot.i_batch - i);

-                gpt_sampler_accept(slot.ctx_sampling, id, true);
+                llama_sampling_accept(slot.ctx_sampling, ctx, id, true);

                slot.n_decoded += 1;
                if (slot.n_decoded == 1)
@@ -1945,14 +1946,19 @@ struct llama_server_context
                    metrics.on_prompt_eval(slot);
                }

+                llama_token_data_array cur_p = { slot.ctx_sampling->cur.data(), slot.ctx_sampling->cur.size(), false };
                result.tok = id;
-                const auto * cur_p = gpt_sampler_get_candidates(slot.ctx_sampling);

-                for (size_t i = 0; i < (size_t) slot.sparams.n_probs; ++i) {
-                    result.probs.push_back({
-                        cur_p->data[i].id,
-                        i >= cur_p->size ? 0.0f : cur_p->data[i].p,
-                    });
+                const int32_t n_probs = slot.sparams.n_probs;
+                if (slot.sparams.temp <= 0 && n_probs > 0)
+                {
+                    // for llama_sample_token_greedy we need to sort candidates
+                    llama_sample_softmax(ctx, &cur_p);
+                }
+
+                for (size_t i = 0; i < std::min(cur_p.size, (size_t)n_probs); ++i)
+                {
+                    result.probs.push_back({cur_p.data[i].id, cur_p.data[i].p});
                }

                if (!process_token(result, slot))
@@ -2204,7 +2210,7 @@ static void params_parse(const backend::ModelOptions* request,
    params.model_alias =  request->modelfile();
    params.n_ctx = request->contextsize();
    //params.memory_f16 = request->f16memory();
-    params.cpuparams.n_threads = request->threads();
+    params.n_threads = request->threads();
    params.n_gpu_layers = request->ngpulayers();
    params.n_batch = request->nbatch();
    // Set params.n_parallel by environment variable (LLAMA_PARALLEL), defaults to 1
--- a/backend/cpp/llama/patches/01-llava.patch
+++ b/backend/cpp/llama/patches/01-llava.patch
@@ -1,13 +0,0 @@
-diff --git a/examples/llava/clip.cpp b/examples/llava/clip.cpp
-index 342042ff..224db9b5 100644
--- a/examples/llava/clip.cpp
-+++ b/examples/llava/clip.cpp
-@@ -2419,7 +2419,7 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
-             struct ggml_tensor * patches = ggml_graph_get_tensor(gf, "patches");
-             int* patches_data = (int*)malloc(ggml_nbytes(patches));
-             for (int i = 0; i < num_patches; i++) {
-                patches_data[i] = i + 1;
-+                patches_data[i] = i;
-             }
-             ggml_backend_tensor_set(patches, patches_data, 0, ggml_nbytes(patches));
-             free(patches_data);
--- a/backend/cpp/llama/prepare.sh
+++ b/backend/cpp/llama/prepare.sh
@@ -1,12 +1,5 @@
 #!/bin/bash

-## Patches
-## Apply patches from the `patches` directory
-for patch in $(ls patches); do
-    echo "Applying patch $patch"
-    patch -d llama.cpp/ -p1 < patches/$patch
-done 
-
 cp -r CMakeLists.txt llama.cpp/examples/grpc-server/
 cp -r grpc-server.cpp llama.cpp/examples/grpc-server/
 cp -rfv json.hpp llama.cpp/examples/grpc-server/
--- a/backend/cpp/llama/utils.hpp
+++ b/backend/cpp/llama/utils.hpp
@@ -480,4 +480,31 @@ static inline std::vector<uint8_t> base64_decode(const std::string & encoded_str
    }

    return ret;
+}
+
+//
+// random string / id
+//
+
+static std::string random_string()
+{
+    static const std::string str("0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz");
+
+    std::random_device rd;
+    std::mt19937 generator(rd());
+
+    std::string result(32, ' ');
+
+    for (int i = 0; i < 32; ++i) {
+        result[i] = str[generator() % str.size()];
+    }
+
+    return result;
+}
+
+static std::string gen_chatcmplid()
+{
+    std::stringstream chatcmplid;
+    chatcmplid << "chatcmpl-" << random_string();
+    return chatcmplid.str();
 }
--- a/backend/go/transcribe/whisper/main.go
+++ b/backend/go/transcribe/whisper/main.go
--- a/backend/go/transcribe/transcript.go
+++ b/backend/go/transcribe/transcript.go
@@ -0,0 +1,104 @@
+package main
+
+import (
+	"fmt"
+	"os"
+	"os/exec"
+	"path/filepath"
+
+	"github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
+	"github.com/go-audio/wav"
+	"github.com/mudler/LocalAI/core/schema"
+)
+
+func ffmpegCommand(args []string) (string, error) {
+	cmd := exec.Command("ffmpeg", args...) // Constrain this to ffmpeg to permit security scanner to see that the command is safe.
+	cmd.Env = os.Environ()
+	out, err := cmd.CombinedOutput()
+	return string(out), err
+}
+
+// AudioToWav converts audio to wav for transcribe.
+// TODO: use https://github.com/mccoyst/ogg?
+func audioToWav(src, dst string) error {
+	commandArgs := []string{"-i", src, "-format", "s16le", "-ar", "16000", "-ac", "1", "-acodec", "pcm_s16le", dst}
+	out, err := ffmpegCommand(commandArgs)
+	if err != nil {
+		return fmt.Errorf("error: %w out: %s", err, out)
+	}
+	return nil
+}
+
+func Transcript(model whisper.Model, audiopath, language string, translate bool, threads uint) (schema.TranscriptionResult, error) {
+	res := schema.TranscriptionResult{}
+
+	dir, err := os.MkdirTemp("", "whisper")
+	if err != nil {
+		return res, err
+	}
+	defer os.RemoveAll(dir)
+
+	convertedPath := filepath.Join(dir, "converted.wav")
+
+	if err := audioToWav(audiopath, convertedPath); err != nil {
+		return res, err
+	}
+
+	// Open samples
+	fh, err := os.Open(convertedPath)
+	if err != nil {
+		return res, err
+	}
+	defer fh.Close()
+
+	// Read samples
+	d := wav.NewDecoder(fh)
+	buf, err := d.FullPCMBuffer()
+	if err != nil {
+		return res, err
+	}
+
+	data := buf.AsFloat32Buffer().Data
+
+	// Process samples
+	context, err := model.NewContext()
+	if err != nil {
+		return res, err
+
+	}
+
+	context.SetThreads(threads)
+
+	if language != "" {
+		context.SetLanguage(language)
+	} else {
+		context.SetLanguage("auto")
+	}
+
+	if translate {
+		context.SetTranslate(true)
+	}
+
+	if err := context.Process(data, nil, nil); err != nil {
+		return res, err
+	}
+
+	for {
+		s, err := context.NextSegment()
+		if err != nil {
+			break
+		}
+
+		var tokens []int
+		for _, t := range s.Tokens {
+			tokens = append(tokens, t.Id)
+		}
+
+		segment := schema.Segment{Id: s.Num, Text: s.Text, Start: s.Start, End: s.End, Tokens: tokens}
+		res.Segments = append(res.Segments, segment)
+
+		res.Text += s.Text
+	}
+
+	return res, nil
+}
--- a/backend/go/transcribe/whisper.go
+++ b/backend/go/transcribe/whisper.go
@@ -0,0 +1,26 @@
+package main
+
+// This is a wrapper to statisfy the GRPC service interface
+// It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)
+import (
+	"github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
+	"github.com/mudler/LocalAI/core/schema"
+	"github.com/mudler/LocalAI/pkg/grpc/base"
+	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
+)
+
+type Whisper struct {
+	base.SingleThread
+	whisper whisper.Model
+}
+
+func (sd *Whisper) Load(opts *pb.ModelOptions) error {
+	// Note: the Model here is a path to a directory containing the model files
+	w, err := whisper.New(opts.ModelFile)
+	sd.whisper = w
+	return err
+}
+
+func (sd *Whisper) AudioTranscription(opts *pb.TranscriptRequest) (schema.TranscriptionResult, error) {
+	return Transcript(sd.whisper, opts.Dst, opts.Language, opts.Translate, uint(opts.Threads))
+}
--- a/backend/go/transcribe/whisper/whisper.go
+++ b/backend/go/transcribe/whisper/whisper.go
@@ -1,105 +0,0 @@
-package main
-
-// This is a wrapper to statisfy the GRPC service interface
-// It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)
-import (
-	"os"
-	"path/filepath"
-
-	"github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
-	"github.com/go-audio/wav"
-	"github.com/mudler/LocalAI/pkg/grpc/base"
-	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
-	"github.com/mudler/LocalAI/pkg/utils"
-)
-
-type Whisper struct {
-	base.SingleThread
-	whisper whisper.Model
-}
-
-func (sd *Whisper) Load(opts *pb.ModelOptions) error {
-	// Note: the Model here is a path to a directory containing the model files
-	w, err := whisper.New(opts.ModelFile)
-	sd.whisper = w
-	return err
-}
-
-func (sd *Whisper) AudioTranscription(opts *pb.TranscriptRequest) (pb.TranscriptResult, error) {
-
-	dir, err := os.MkdirTemp("", "whisper")
-	if err != nil {
-		return pb.TranscriptResult{}, err
-	}
-	defer os.RemoveAll(dir)
-
-	convertedPath := filepath.Join(dir, "converted.wav")
-
-	if err := utils.AudioToWav(opts.Dst, convertedPath); err != nil {
-		return pb.TranscriptResult{}, err
-	}
-
-	// Open samples
-	fh, err := os.Open(convertedPath)
-	if err != nil {
-		return pb.TranscriptResult{}, err
-	}
-	defer fh.Close()
-
-	// Read samples
-	d := wav.NewDecoder(fh)
-	buf, err := d.FullPCMBuffer()
-	if err != nil {
-		return pb.TranscriptResult{}, err
-	}
-
-	data := buf.AsFloat32Buffer().Data
-
-	// Process samples
-	context, err := sd.whisper.NewContext()
-	if err != nil {
-		return pb.TranscriptResult{}, err
-
-	}
-
-	context.SetThreads(uint(opts.Threads))
-
-	if opts.Language != "" {
-		context.SetLanguage(opts.Language)
-	} else {
-		context.SetLanguage("auto")
-	}
-
-	if opts.Translate {
-		context.SetTranslate(true)
-	}
-
-	if err := context.Process(data, nil, nil); err != nil {
-		return pb.TranscriptResult{}, err
-	}
-
-	segments := []*pb.TranscriptSegment{}
-	text := ""
-	for {
-		s, err := context.NextSegment()
-		if err != nil {
-			break
-		}
-
-		var tokens []int32
-		for _, t := range s.Tokens {
-			tokens = append(tokens, int32(t.Id))
-		}
-
-		segment := &pb.TranscriptSegment{Id: int32(s.Num), Text: s.Text, Start: int64(s.Start), End: int64(s.End), Tokens: tokens}
-		segments = append(segments, segment)
-
-		text += s.Text
-	}
-
-	return pb.TranscriptResult{
-		Segments: segments,
-		Text:     text,
-	}, nil
-
-}
--- a/backend/python/autogptq/requirements-intel.txt
+++ b/backend/python/autogptq/requirements-intel.txt
@@ -2,4 +2,4 @@
 intel-extension-for-pytorch
 torch
 optimum[openvino]
-setuptools==75.1.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406
--- a/backend/python/autogptq/requirements.txt
+++ b/backend/python/autogptq/requirements.txt
@@ -1,6 +1,6 @@
 accelerate
 auto-gptq==0.7.1
-grpcio==1.66.1
+grpcio==1.65.4
 protobuf
 certifi
 transformers
--- a/backend/python/bark/requirements-intel.txt
+++ b/backend/python/bark/requirements-intel.txt
@@ -3,6 +3,6 @@ intel-extension-for-pytorch
 torch
 torchaudio
 optimum[openvino]
-setuptools==75.1.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
 transformers
 accelerate
--- a/backend/python/bark/requirements.txt
+++ b/backend/python/bark/requirements.txt
@@ -1,4 +1,4 @@
 bark==0.1.5
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
 certifi
--- a/backend/python/common/template/requirements.txt
+++ b/backend/python/common/template/requirements.txt
@@ -1,2 +1,2 @@
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
--- a/backend/python/coqui/requirements-intel.txt
+++ b/backend/python/coqui/requirements-intel.txt
@@ -3,6 +3,6 @@ intel-extension-for-pytorch
 torch
 torchaudio
 optimum[openvino]
-setuptools==75.1.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406
 transformers
 accelerate
--- a/backend/python/coqui/requirements.txt
+++ b/backend/python/coqui/requirements.txt
@@ -1,4 +1,4 @@
-coqui-tts
-grpcio==1.66.1
+TTS==0.22.0
+grpcio==1.65.5
 protobuf
 certifi
--- a/backend/python/diffusers/backend.py
+++ b/backend/python/diffusers/backend.py
@@ -168,7 +168,7 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
            if request.CFGScale != 0:
                self.cfg_scale = request.CFGScale

-            clipmodel = "Lykon/dreamshaper-8"
+            clipmodel = "runwayml/stable-diffusion-v1-5"
            if request.CLIPModel != "":
                clipmodel = request.CLIPModel
            clipsubfolder = "text_encoder"
--- a/backend/python/diffusers/requirements-intel.txt
+++ b/backend/python/diffusers/requirements-intel.txt
@@ -3,7 +3,7 @@ intel-extension-for-pytorch
 torch
 torchvision
 optimum[openvino]
-setuptools==75.1.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
 diffusers
 opencv-python
 transformers
--- a/backend/python/diffusers/requirements.txt
+++ b/backend/python/diffusers/requirements.txt
@@ -1,5 +1,5 @@
 setuptools
-grpcio==1.66.1
+grpcio==1.65.4
 pillow
 protobuf
 certifi
--- a/backend/python/diffusers/test.py
+++ b/backend/python/diffusers/test.py
@@ -53,7 +53,7 @@ class TestBackendServicer(unittest.TestCase):
            self.setUp()
            with grpc.insecure_channel("localhost:50051") as channel:
                stub = backend_pb2_grpc.BackendStub(channel)
-                response = stub.LoadModel(backend_pb2.ModelOptions(Model="Lykon/dreamshaper-8"))
+                response = stub.LoadModel(backend_pb2.ModelOptions(Model="runwayml/stable-diffusion-v1-5"))
                self.assertTrue(response.success)
                self.assertEqual(response.message, "Model loaded successfully")
        except Exception as err:
@@ -71,7 +71,7 @@ class TestBackendServicer(unittest.TestCase):
            self.setUp()
            with grpc.insecure_channel("localhost:50051") as channel:
                stub = backend_pb2_grpc.BackendStub(channel)
-                response = stub.LoadModel(backend_pb2.ModelOptions(Model="Lykon/dreamshaper-8"))
+                response = stub.LoadModel(backend_pb2.ModelOptions(Model="runwayml/stable-diffusion-v1-5"))
                print(response.message)
                self.assertTrue(response.success)
                image_req = backend_pb2.GenerateImageRequest(positive_prompt="cat", width=16,height=16, dst="test.jpg")
@@ -81,4 +81,4 @@ class TestBackendServicer(unittest.TestCase):
            print(err)
            self.fail("Image gen service failed")
        finally:
-            self.tearDown()
+            self.tearDown()
--- a/backend/python/exllama/.gitignore
+++ b/backend/python/exllama/.gitignore
@@ -0,0 +1 @@
+source
--- a/backend/python/exllama/Makefile
+++ b/backend/python/exllama/Makefile
@@ -0,0 +1,25 @@
+export CONDA_ENV_PATH = "exllama.yml"
+
+.PHONY: exllama
+exllama: protogen
+	bash install.sh ${CONDA_ENV_PATH}
+
+.PHONY: run
+run: protogen
+	@echo "Running exllama..."
+	bash run.sh
+	@echo "exllama run."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
+
+.PHONY: clean
+clean: protogen-clean
+	$(RM) -r venv source __pycache__
--- a/backend/python/exllama/README.md
+++ b/backend/python/exllama/README.md
@@ -0,0 +1,5 @@
+# Creating a separate environment for the exllama project
+
+```
+make exllama
+```
--- a/backend/python/exllama/backend.py
+++ b/backend/python/exllama/backend.py
@@ -0,0 +1,159 @@
+#!/usr/bin/env python3
+import grpc
+from concurrent import futures
+import time
+import backend_pb2
+import backend_pb2_grpc
+import argparse
+import signal
+import sys
+import os, glob
+
+from pathlib import Path
+import torch
+import torch.nn.functional as F
+from torch import version as torch_version
+
+from source.tokenizer import ExLlamaTokenizer
+from source.generator import ExLlamaGenerator
+from source.model import ExLlama, ExLlamaCache, ExLlamaConfig
+
+_ONE_DAY_IN_SECONDS = 60 * 60 * 24
+
+# If MAX_WORKERS are specified in the environment use it, otherwise default to 1
+MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
+
+# Implement the BackendServicer class with the service methods
+class BackendServicer(backend_pb2_grpc.BackendServicer):
+    def generate(self,prompt, max_new_tokens):
+        self.generator.end_beam_search()
+
+        # Tokenizing the input
+        ids = self.generator.tokenizer.encode(prompt)
+
+        self.generator.gen_begin_reuse(ids)
+        initial_len = self.generator.sequence[0].shape[0]
+        has_leading_space = False
+        decoded_text = ''
+        for i in range(max_new_tokens):
+            token = self.generator.gen_single_token()
+            if i == 0 and self.generator.tokenizer.tokenizer.IdToPiece(int(token)).startswith('▁'):
+                has_leading_space = True
+
+            decoded_text = self.generator.tokenizer.decode(self.generator.sequence[0][initial_len:])
+            if has_leading_space:
+                decoded_text = ' ' + decoded_text
+
+            if token.item() == self.generator.tokenizer.eos_token_id:
+                break
+        return decoded_text
+    def Health(self, request, context):
+        return backend_pb2.Reply(message=bytes("OK", 'utf-8'))
+    def LoadModel(self, request, context):
+        try:
+            # https://github.com/turboderp/exllama/blob/master/example_cfg.py
+            model_directory = request.ModelFile
+
+            # Locate files we need within that directory
+            tokenizer_path = os.path.join(model_directory, "tokenizer.model")
+            model_config_path = os.path.join(model_directory, "config.json")
+            st_pattern = os.path.join(model_directory, "*.safetensors")
+            model_path = glob.glob(st_pattern)[0]
+
+            # Create config, model, tokenizer and generator
+
+            config = ExLlamaConfig(model_config_path)               # create config from config.json
+            config.model_path = model_path                          # supply path to model weights file
+            if (request.ContextSize):
+                config.max_seq_len = request.ContextSize            # override max sequence length
+                config.max_attention_size = request.ContextSize**2  # Should be set to context_size^2. 
+                # https://github.com/turboderp/exllama/issues/220#issuecomment-1720324163
+
+            # Set Rope scaling.
+            if (request.RopeFreqScale):
+                # Alpha value for Rope scaling. 
+                # Higher value increases context but adds perplexity.
+                # alpha_value and compress_pos_emb are mutually exclusive.
+                # https://github.com/turboderp/exllama/issues/115
+                config.alpha_value = request.RopeFreqScale
+                config.calculate_rotary_embedding_base()
+
+            model = ExLlama(config)                                 # create ExLlama instance and load the weights
+            tokenizer = ExLlamaTokenizer(tokenizer_path)            # create tokenizer from tokenizer model file
+
+            cache = ExLlamaCache(model, batch_size = 2)             # create cache for inference
+            generator = ExLlamaGenerator(model, tokenizer, cache)   # create generator
+
+            self.generator= generator
+            self.model = model
+            self.tokenizer = tokenizer
+            self.cache = cache
+        except Exception as err:
+            return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
+        return backend_pb2.Result(message="Model loaded successfully", success=True)
+
+    def Predict(self, request, context):
+        penalty = 1.15
+        if request.Penalty != 0.0:
+            penalty = request.Penalty
+        self.generator.settings.token_repetition_penalty_max = penalty
+        self.generator.settings.temperature = request.Temperature
+        self.generator.settings.top_k = request.TopK
+        self.generator.settings.top_p = request.TopP
+
+        tokens = 512
+        if request.Tokens != 0:
+            tokens = request.Tokens
+
+        if self.cache.batch_size == 1:
+            del self.cache
+            self.cache = ExLlamaCache(self.model, batch_size=2)
+            self.generator = ExLlamaGenerator(self.model, self.tokenizer, self.cache)
+
+        t = self.generate(request.Prompt, tokens)
+
+        # Remove prompt from response if present
+        if request.Prompt in t:
+            t = t.replace(request.Prompt, "")
+
+        return backend_pb2.Result(message=bytes(t, encoding='utf-8'))
+
+    def PredictStream(self, request, context):
+        # Implement PredictStream RPC
+        #for reply in some_data_generator():
+        #    yield reply
+        # Not implemented yet
+        return self.Predict(request, context)
+
+
+def serve(address):
+    server = grpc.server(futures.ThreadPoolExecutor(max_workers=MAX_WORKERS))
+    backend_pb2_grpc.add_BackendServicer_to_server(BackendServicer(), server)
+    server.add_insecure_port(address)
+    server.start()
+    print("Server started. Listening on: " + address, file=sys.stderr)
+
+    # Define the signal handler function
+    def signal_handler(sig, frame):
+        print("Received termination signal. Shutting down...")
+        server.stop(0)
+        sys.exit(0)
+
+    # Set the signal handlers for SIGINT and SIGTERM
+    signal.signal(signal.SIGINT, signal_handler)
+    signal.signal(signal.SIGTERM, signal_handler)
+
+    try:
+        while True:
+            time.sleep(_ONE_DAY_IN_SECONDS)
+    except KeyboardInterrupt:
+        server.stop(0)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Run the gRPC server.")
+    parser.add_argument(
+        "--addr", default="localhost:50051", help="The address to bind the server to."
+    )
+    args = parser.parse_args()
+
+    serve(args.addr)
--- a/backend/python/exllama/install.sh
+++ b/backend/python/exllama/install.sh
@@ -0,0 +1,13 @@
+#!/bin/bash
+set -e
+
+LIMIT_TARGETS="cublas"
+
+source $(dirname $0)/../common/libbackend.sh
+
+installRequirements
+
+git clone https://github.com/turboderp/exllama $MY_DIR/source
+uv pip install ${BUILD_ISOLATION_FLAG} --requirement ${MY_DIR}/source/requirements.txt
+
+cp -v ./*py $MY_DIR/source/
--- a/backend/python/exllama/requirements-cpu.txt
+++ b/backend/python/exllama/requirements-cpu.txt
@@ -0,0 +1,3 @@
+transformers
+accelerate
+torch
--- a/backend/python/exllama/requirements-cublas11.txt
+++ b/backend/python/exllama/requirements-cublas11.txt
@@ -0,0 +1,4 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch
+transformers
+accelerate
--- a/backend/python/exllama/requirements-cublas12.txt
+++ b/backend/python/exllama/requirements-cublas12.txt
@@ -0,0 +1,3 @@
+torch
+transformers
+accelerate
--- a/backend/python/exllama/requirements.txt
+++ b/backend/python/exllama/requirements.txt
@@ -0,0 +1,4 @@
+grpcio==1.65.5
+protobuf
+certifi
+setuptools
--- a/backend/python/exllama/run.sh
+++ b/backend/python/exllama/run.sh
@@ -0,0 +1,7 @@
+#!/bin/bash
+LIMIT_TARGETS="cublas"
+BACKEND_FILE="${MY_DIR}/source/backend.py"
+
+source $(dirname $0)/../common/libbackend.sh
+
+startBackend $@
--- a/backend/python/exllama/test.sh
+++ b/backend/python/exllama/test.sh
@@ -0,0 +1,6 @@
+#!/bin/bash
+set -e
+
+source $(dirname $0)/../common/libbackend.sh
+
+runUnittests
--- a/backend/python/exllama2/requirements.txt
+++ b/backend/python/exllama2/requirements.txt
@@ -1,4 +1,4 @@
-grpcio==1.66.1
+grpcio==1.65.4
 protobuf
 certifi
 wheel
--- a/backend/python/mamba/requirements.txt
+++ b/backend/python/mamba/requirements.txt
@@ -1,3 +1,3 @@
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
 certifi
--- a/backend/python/openvoice/requirements-intel.txt
+++ b/backend/python/openvoice/requirements-intel.txt
@@ -2,7 +2,7 @@
 intel-extension-for-pytorch
 torch
 optimum[openvino]
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
 librosa==0.9.1
 faster-whisper==1.0.3
--- a/backend/python/openvoice/requirements.txt
+++ b/backend/python/openvoice/requirements.txt
@@ -1,4 +1,4 @@
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
 librosa
 faster-whisper
--- a/backend/python/parler-tts/install.sh
+++ b/backend/python/parler-tts/install.sh
@@ -15,12 +15,5 @@ installRequirements

 # https://github.com/descriptinc/audiotools/issues/101
 # incompatible protobuf versions.
-PYDIR=python3.10
-pyenv="${MY_DIR}/venv/lib/${PYDIR}/site-packages/google/protobuf/internal/"
-
-if [ ! -d ${pyenv} ]; then
-    echo "(parler-tts/install.sh): Error: ${pyenv} does not exist"
-    exit 1
-fi
-
-curl -L https://raw.githubusercontent.com/protocolbuffers/protobuf/main/python/google/protobuf/internal/builder.py -o ${pyenv}/builder.py
+PYDIR=$(ls ${MY_DIR}/venv/lib)
+curl -L https://raw.githubusercontent.com/protocolbuffers/protobuf/main/python/google/protobuf/internal/builder.py -o ${MY_DIR}/venv/lib/${PYDIR}/site-packages/google/protobuf/internal/builder.py
--- a/backend/python/parler-tts/requirements-hipblas.txt
+++ b/backend/python/parler-tts/requirements-hipblas.txt
@@ -1,5 +1,5 @@
 --extra-index-url https://download.pytorch.org/whl/rocm6.0
-torch==2.3.0+rocm6.0
-torchaudio==2.3.0+rocm6.0
+torch
+torchaudio
 transformers
-accelerate
+accelerate
--- a/backend/python/parler-tts/requirements-intel.txt
+++ b/backend/python/parler-tts/requirements-intel.txt
@@ -3,6 +3,6 @@ intel-extension-for-pytorch
 torch
 torchaudio
 optimum[openvino]
-setuptools==75.1.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406
 transformers
 accelerate
--- a/backend/python/parler-tts/requirements.txt
+++ b/backend/python/parler-tts/requirements.txt
@@ -1,4 +1,4 @@
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
 certifi
 llvmlite==0.43.0
--- a/backend/python/rerankers/requirements-intel.txt
+++ b/backend/python/rerankers/requirements-intel.txt
@@ -5,4 +5,4 @@ accelerate
 torch
 rerankers[transformers]
 optimum[openvino]
-setuptools==75.1.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406
--- a/backend/python/rerankers/requirements.txt
+++ b/backend/python/rerankers/requirements.txt
@@ -1,3 +1,3 @@
-grpcio==1.66.1
+grpcio==1.65.4
 protobuf
 certifi
--- a/backend/python/sentencetransformers/backend.py
+++ b/backend/python/sentencetransformers/backend.py
@@ -55,7 +55,7 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
        """
        model_name = request.Model
        try:
-            self.model = SentenceTransformer(model_name, trust_remote_code=request.TrustRemoteCode)
+            self.model = SentenceTransformer(model_name)
        except Exception as err:
            return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")

--- a/backend/python/sentencetransformers/requirements-cpu.txt
+++ b/backend/python/sentencetransformers/requirements-cpu.txt
@@ -2,5 +2,5 @@ torch
 accelerate
 transformers
 bitsandbytes
-sentence-transformers==3.1.0
+sentence-transformers==3.0.1
 transformers
--- a/backend/python/sentencetransformers/requirements-cublas11.txt
+++ b/backend/python/sentencetransformers/requirements-cublas11.txt
@@ -1,5 +1,5 @@
 --extra-index-url https://download.pytorch.org/whl/cu118
 torch
 accelerate
-sentence-transformers==3.1.0
+sentence-transformers==3.0.1
 transformers
--- a/backend/python/sentencetransformers/requirements-cublas12.txt
+++ b/backend/python/sentencetransformers/requirements-cublas12.txt
@@ -1,4 +1,4 @@
 torch
 accelerate
-sentence-transformers==3.1.0
+sentence-transformers==3.0.1
 transformers
--- a/backend/python/sentencetransformers/requirements-hipblas.txt
+++ b/backend/python/sentencetransformers/requirements-hipblas.txt
@@ -1,5 +1,5 @@
 --extra-index-url https://download.pytorch.org/whl/rocm6.0
 torch
 accelerate
-sentence-transformers==3.1.0
+sentence-transformers==3.0.1
 transformers
--- a/backend/python/sentencetransformers/requirements-intel.txt
+++ b/backend/python/sentencetransformers/requirements-intel.txt
@@ -4,5 +4,5 @@ torch
 optimum[openvino]
 setuptools==69.5.1 # https://github.com/mudler/LocalAI/issues/2406
 accelerate
-sentence-transformers==3.1.0
+sentence-transformers==3.0.1
 transformers
--- a/backend/python/sentencetransformers/requirements.txt
+++ b/backend/python/sentencetransformers/requirements.txt
@@ -1,5 +1,3 @@
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
-certifi
-datasets
-einops
+certifi
--- a/backend/python/transformers-musicgen/backend.py
+++ b/backend/python/transformers-musicgen/backend.py
@@ -15,7 +15,7 @@ import backend_pb2_grpc

 import grpc

-from scipy.io import wavfile
+from scipy.io.wavfile import write as write_wav
 from transformers import AutoProcessor, MusicgenForConditionalGeneration

 _ONE_DAY_IN_SECONDS = 60 * 60 * 24
@@ -63,61 +63,6 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):

        return backend_pb2.Result(message="Model loaded successfully", success=True)

-    def SoundGeneration(self, request, context):
-        model_name = request.model
-        if model_name == "":
-            return backend_pb2.Result(success=False, message="request.model is required")
-        try:
-            self.processor = AutoProcessor.from_pretrained(model_name)
-            self.model = MusicgenForConditionalGeneration.from_pretrained(model_name)
-            inputs = None
-            if request.text == "":
-                inputs = self.model.get_unconditional_inputs(num_samples=1)
-            elif request.HasField('src'):
-                # TODO SECURITY CODE GOES HERE LOL
-                # WHO KNOWS IF THIS WORKS???
-                sample_rate, wsamples = wavfile.read('path_to_your_file.wav')
-                
-                if request.HasField('src_divisor'):
-                    wsamples = wsamples[: len(wsamples) // request.src_divisor]
-                
-                inputs = self.processor(
-                    audio=wsamples,
-                    sampling_rate=sample_rate,
-                    text=[request.text],
-                    padding=True,
-                    return_tensors="pt",
-                )
-            else:
-                inputs = self.processor(
-                    text=[request.text],
-                    padding=True,
-                    return_tensors="pt",
-                )
-            
-            tokens = 256
-            if request.HasField('duration'):
-                tokens = int(request.duration * 51.2) # 256 tokens = 5 seconds, therefore 51.2 tokens is one second
-            guidance = 3.0
-            if request.HasField('temperature'):
-                guidance = request.temperature
-            dosample = True
-            if request.HasField('sample'):
-                dosample = request.sample
-            audio_values = self.model.generate(**inputs, do_sample=dosample, guidance_scale=guidance, max_new_tokens=tokens)
-            print("[transformers-musicgen] SoundGeneration generated!", file=sys.stderr)
-            sampling_rate = self.model.config.audio_encoder.sampling_rate
-            wavfile.write(request.dst, rate=sampling_rate, data=audio_values[0, 0].numpy())
-            print("[transformers-musicgen] SoundGeneration saved to", request.dst, file=sys.stderr)
-            print("[transformers-musicgen] SoundGeneration for", file=sys.stderr)
-            print("[transformers-musicgen] SoundGeneration requested tokens", tokens, file=sys.stderr)
-            print(request, file=sys.stderr)
-        except Exception as err:
-            return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
-        return backend_pb2.Result(success=True)
-
-
-# The TTS endpoint is older, and provides fewer features, but exists for compatibility reasons
    def TTS(self, request, context):
        model_name = request.model
        if model_name == "":
@@ -130,7 +75,8 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
                padding=True,
                return_tensors="pt",
            )
-            tokens = 512 # No good place to set the "length" in TTS, so use 10s as a sane default
+            tokens = 256
+            # TODO get tokens from request?
            audio_values = self.model.generate(**inputs, max_new_tokens=tokens)
            print("[transformers-musicgen] TTS generated!", file=sys.stderr)
            sampling_rate = self.model.config.audio_encoder.sampling_rate
--- a/backend/python/transformers-musicgen/requirements-intel.txt
+++ b/backend/python/transformers-musicgen/requirements-intel.txt
@@ -4,4 +4,4 @@ transformers
 accelerate
 torch
 optimum[openvino]
-setuptools==75.1.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==69.5.1 # https://github.com/mudler/LocalAI/issues/2406
--- a/backend/python/transformers-musicgen/requirements.txt
+++ b/backend/python/transformers-musicgen/requirements.txt
@@ -1,4 +1,4 @@
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
 scipy==1.14.0
 certifi
--- a/backend/python/transformers-musicgen/test.py
+++ b/backend/python/transformers-musicgen/test.py
@@ -63,7 +63,7 @@ class TestBackendServicer(unittest.TestCase):

    def test_tts(self):
        """
-        This method tests if TTS is generated successfully
+        This method tests if the embeddings are generated successfully
        """
        try:
            self.setUp()
@@ -77,24 +77,5 @@ class TestBackendServicer(unittest.TestCase):
        except Exception as err:
            print(err)
            self.fail("TTS service failed")
-        finally:
-            self.tearDown()
-
-    def test_sound_generation(self):
-        """
-        This method tests if SoundGeneration is generated successfully
-        """
-        try:
-            self.setUp()
-            with grpc.insecure_channel("localhost:50051") as channel:
-                stub = backend_pb2_grpc.BackendStub(channel)
-                response = stub.LoadModel(backend_pb2.ModelOptions(Model="facebook/musicgen-small"))
-                self.assertTrue(response.success)
-                sg_request = backend_pb2.SoundGenerationRequest(text="80s TV news production music hit for tonight's biggest story")
-                sg_response = stub.SoundGeneration(sg_request)
-                self.assertIsNotNone(sg_response)
-        except Exception as err:
-            print(err)
-            self.fail("SoundGeneration service failed")
        finally:
            self.tearDown()
--- a/backend/python/transformers/requirements.txt
+++ b/backend/python/transformers/requirements.txt
@@ -1,4 +1,4 @@
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
 certifi
 setuptools==69.5.1 # https://github.com/mudler/LocalAI/issues/2406
--- a/backend/python/vall-e-x/requirements-intel.txt
+++ b/backend/python/vall-e-x/requirements-intel.txt
@@ -4,4 +4,4 @@ accelerate
 torch
 torchaudio
 optimum[openvino]
-setuptools==75.1.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406
--- a/backend/python/vall-e-x/requirements.txt
+++ b/backend/python/vall-e-x/requirements.txt
@@ -1,3 +1,3 @@
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
 certifi
--- a/backend/python/vllm/backend.py
+++ b/backend/python/vllm/backend.py
@@ -135,26 +135,6 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
        res = await gen.__anext__()
        return res

-    def Embedding(self, request, context):
-        """
-        A gRPC method that calculates embeddings for a given sentence.
-
-        Args:
-            request: An EmbeddingRequest object that contains the request parameters.
-            context: A grpc.ServicerContext object that provides information about the RPC.
-
-        Returns:
-            An EmbeddingResult object that contains the calculated embeddings.
-        """
-        print("Calculated embeddings for: " + request.Embeddings, file=sys.stderr)
-        outputs = self.model.encode(request.Embeddings)
-        # Check if we have one result at least
-        if len(outputs) == 0:
-            context.set_code(grpc.StatusCode.INVALID_ARGUMENT)
-            context.set_details("No embeddings were calculated.")
-            return backend_pb2.EmbeddingResult()
-        return backend_pb2.EmbeddingResult(embeddings=outputs[0].outputs.embedding)
-
    async def PredictStream(self, request, context):
        """
        Generates text based on the given prompt and sampling parameters, and streams the results.
--- a/backend/python/vllm/requirements-intel.txt
+++ b/backend/python/vllm/requirements-intel.txt
@@ -4,4 +4,4 @@ accelerate
 torch
 transformers
 optimum[openvino]
-setuptools==75.1.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
--- a/backend/python/vllm/requirements.txt
+++ b/backend/python/vllm/requirements.txt
@@ -1,4 +1,4 @@
-grpcio==1.66.1
+grpcio==1.65.5
 protobuf
 certifi
 setuptools
--- a/backend/python/vllm/test.py
+++ b/backend/python/vllm/test.py
@@ -72,28 +72,5 @@ class TestBackendServicer(unittest.TestCase):
        except Exception as err:
            print(err)
            self.fail("text service failed")
-        finally:
-            self.tearDown()
-
-    def test_embedding(self):
-        """
-        This method tests if the embeddings are generated successfully
-        """
-        try:
-            self.setUp()
-            with grpc.insecure_channel("localhost:50051") as channel:
-                stub = backend_pb2_grpc.BackendStub(channel)
-                response = stub.LoadModel(backend_pb2.ModelOptions(Model="intfloat/e5-mistral-7b-instruct"))
-                self.assertTrue(response.success)
-                embedding_request = backend_pb2.PredictOptions(Embeddings="This is a test sentence.")
-                embedding_response = stub.Embedding(embedding_request)
-                self.assertIsNotNone(embedding_response.embeddings)
-                # assert that is a list of floats
-                self.assertIsInstance(embedding_response.embeddings, list)
-                # assert that the list is not empty
-                self.assertTrue(len(embedding_response.embeddings) > 0)
-        except Exception as err:
-            print(err)
-            self.fail("Embedding service failed")
        finally:
            self.tearDown()
--- a/core/backend/backend_suite_test.go
+++ b/core/backend/backend_suite_test.go
@@ -1,13 +0,0 @@
-package backend_test
-
-import (
-	"testing"
-
-	. "github.com/onsi/ginkgo/v2"
-	. "github.com/onsi/gomega"
-)
-
-func TestBackend(t *testing.T) {
-	RegisterFailHandler(Fail)
-	RunSpecs(t, "Backend test suite")
-}
--- a/core/backend/embeddings.go
+++ b/core/backend/embeddings.go
@@ -12,7 +12,7 @@ import (
 func ModelEmbedding(s string, tokens []int, loader *model.ModelLoader, backendConfig config.BackendConfig, appConfig *config.ApplicationConfig) (func() ([]float32, error), error) {
 	modelFile := backendConfig.Model

-	grpcOpts := GRPCModelOpts(backendConfig)
+	grpcOpts := gRPCModelOpts(backendConfig)

 	var inferenceModel interface{}
 	var err error
--- a/core/backend/image.go
+++ b/core/backend/image.go
@@ -12,7 +12,7 @@ func ImageGeneration(height, width, mode, step, seed int, positive_prompt, negat
 	if *threads == 0 && appConfig.Threads != 0 {
 		threads = &appConfig.Threads
 	}
-	gRPCOpts := GRPCModelOpts(backendConfig)
+	gRPCOpts := gRPCModelOpts(backendConfig)
 	opts := modelOpts(backendConfig, appConfig, []model.Option{
 		model.WithBackendString(backendConfig.Backend),
 		model.WithAssetDir(appConfig.AssetsDestination),
--- a/core/backend/llm.go
+++ b/core/backend/llm.go
@@ -9,8 +9,6 @@ import (
 	"sync"
 	"unicode/utf8"

-	"github.com/rs/zerolog/log"
-
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/schema"

@@ -31,13 +29,13 @@ type TokenUsage struct {
 	Completion int
 }

-func ModelInference(ctx context.Context, s string, messages []schema.Message, images, videos, audios []string, loader *model.ModelLoader, c config.BackendConfig, o *config.ApplicationConfig, tokenCallback func(string, TokenUsage) bool) (func() (LLMResponse, error), error) {
+func ModelInference(ctx context.Context, s string, messages []schema.Message, images []string, loader *model.ModelLoader, c config.BackendConfig, o *config.ApplicationConfig, tokenCallback func(string, TokenUsage) bool) (func() (LLMResponse, error), error) {
 	modelFile := c.Model
 	threads := c.Threads
 	if *threads == 0 && o.Threads != 0 {
 		threads = &o.Threads
 	}
-	grpcOpts := GRPCModelOpts(c)
+	grpcOpts := gRPCModelOpts(c)

 	var inferenceModel grpc.Backend
 	var err error
@@ -89,7 +87,7 @@ func ModelInference(ctx context.Context, s string, messages []schema.Message, im
 			case string:
 				protoMessages[i].Content = ct
 			default:
-				return nil, fmt.Errorf("unsupported type for schema.Message.Content for inference: %T", ct)
+				return nil, fmt.Errorf("Unsupported type for schema.Message.Content for inference: %T", ct)
 			}
 		}
 	}
@@ -101,8 +99,6 @@ func ModelInference(ctx context.Context, s string, messages []schema.Message, im
 		opts.Messages = protoMessages
 		opts.UseTokenizerTemplate = c.TemplateConfig.UseTokenizerTemplate
 		opts.Images = images
-		opts.Videos = videos
-		opts.Audios = audios

 		tokenUsage := TokenUsage{}

@@ -185,37 +181,13 @@ func Finetune(config config.BackendConfig, input, prediction string) string {
 		mu.Lock()
 		reg, ok := cutstrings[c]
 		if !ok {
-			r, err := regexp.Compile(c)
-			if err != nil {
-				log.Fatal().Err(err).Msg("failed to compile regex")
-			}
-			cutstrings[c] = r
+			cutstrings[c] = regexp.MustCompile(c)
 			reg = cutstrings[c]
 		}
 		mu.Unlock()
 		prediction = reg.ReplaceAllString(prediction, "")
 	}

-	// extract results from the response which can be for instance inside XML tags
-	var predResult string
-	for _, r := range config.ExtractRegex {
-		mu.Lock()
-		reg, ok := cutstrings[r]
-		if !ok {
-			regex, err := regexp.Compile(r)
-			if err != nil {
-				log.Fatal().Err(err).Msg("failed to compile regex")
-			}
-			cutstrings[r] = regex
-			reg = regex
-		}
-		mu.Unlock()
-		predResult += reg.FindString(prediction)
-	}
-	if predResult != "" {
-		prediction = predResult
-	}
-
 	for _, c := range config.TrimSpace {
 		prediction = strings.TrimSpace(strings.TrimPrefix(prediction, c))
 	}
--- a/core/backend/llm_test.go
+++ b/core/backend/llm_test.go
@@ -1,109 +0,0 @@
-package backend_test
-
-import (
-	. "github.com/mudler/LocalAI/core/backend"
-	"github.com/mudler/LocalAI/core/config"
-	"github.com/mudler/LocalAI/core/schema"
-
-	. "github.com/onsi/ginkgo/v2"
-	. "github.com/onsi/gomega"
-)
-
-var _ = Describe("LLM tests", func() {
-	Context("Finetune LLM output", func() {
-		var (
-			testConfig config.BackendConfig
-			input      string
-			prediction string
-			result     string
-		)
-
-		BeforeEach(func() {
-			testConfig = config.BackendConfig{
-				PredictionOptions: schema.PredictionOptions{
-					Echo: false,
-				},
-				LLMConfig: config.LLMConfig{
-					Cutstrings:   []string{`<.*?>`},                  // Example regex for removing XML tags
-					ExtractRegex: []string{`<result>(.*?)</result>`}, // Example regex to extract from tags
-					TrimSpace:    []string{" ", "\n"},
-					TrimSuffix:   []string{".", "!"},
-				},
-			}
-		})
-
-		Context("when echo is enabled", func() {
-			BeforeEach(func() {
-				testConfig.Echo = true
-				input = "Hello"
-				prediction = "World"
-			})
-
-			It("should prepend input to prediction", func() {
-				result = Finetune(testConfig, input, prediction)
-				Expect(result).To(Equal("HelloWorld"))
-			})
-		})
-
-		Context("when echo is disabled", func() {
-			BeforeEach(func() {
-				testConfig.Echo = false
-				input = "Hello"
-				prediction = "World"
-			})
-
-			It("should not modify the prediction with input", func() {
-				result = Finetune(testConfig, input, prediction)
-				Expect(result).To(Equal("World"))
-			})
-		})
-
-		Context("when cutstrings regex is applied", func() {
-			BeforeEach(func() {
-				input = ""
-				prediction = "<div>Hello</div> World"
-			})
-
-			It("should remove substrings matching cutstrings regex", func() {
-				result = Finetune(testConfig, input, prediction)
-				Expect(result).To(Equal("Hello World"))
-			})
-		})
-
-		Context("when extract regex is applied", func() {
-			BeforeEach(func() {
-				input = ""
-				prediction = "<response><result>42</result></response>"
-			})
-
-			It("should extract substrings matching the extract regex", func() {
-				result = Finetune(testConfig, input, prediction)
-				Expect(result).To(Equal("42"))
-			})
-		})
-
-		Context("when trimming spaces", func() {
-			BeforeEach(func() {
-				input = ""
-				prediction = "   Hello World   "
-			})
-
-			It("should trim spaces from the prediction", func() {
-				result = Finetune(testConfig, input, prediction)
-				Expect(result).To(Equal("Hello World"))
-			})
-		})
-
-		Context("when trimming suffixes", func() {
-			BeforeEach(func() {
-				input = ""
-				prediction = "Hello World."
-			})
-
-			It("should trim suffixes from the prediction", func() {
-				result = Finetune(testConfig, input, prediction)
-				Expect(result).To(Equal("Hello World"))
-			})
-		})
-	})
-})
--- a/core/backend/options.go
+++ b/core/backend/options.go
@@ -44,7 +44,7 @@ func getSeed(c config.BackendConfig) int32 {
 	return seed
 }

-func GRPCModelOpts(c config.BackendConfig) *pb.ModelOptions {
+func gRPCModelOpts(c config.BackendConfig) *pb.ModelOptions {
 	b := 512
 	if c.Batch != 0 {
 		b = c.Batch
--- a/core/backend/rerank.go
+++ b/core/backend/rerank.go
@@ -15,7 +15,7 @@ func Rerank(backend, modelFile string, request *proto.RerankRequest, loader *mod
 		return nil, fmt.Errorf("backend is required")
 	}

-	grpcOpts := GRPCModelOpts(backendConfig)
+	grpcOpts := gRPCModelOpts(backendConfig)

 	opts := modelOpts(config.BackendConfig{}, appConfig, []model.Option{
 		model.WithBackendString(bb),
--- a/core/backend/soundgeneration.go
+++ b/core/backend/soundgeneration.go
@@ -1,74 +0,0 @@
-package backend
-
-import (
-	"context"
-	"fmt"
-	"os"
-	"path/filepath"
-
-	"github.com/mudler/LocalAI/core/config"
-	"github.com/mudler/LocalAI/pkg/grpc/proto"
-	"github.com/mudler/LocalAI/pkg/model"
-	"github.com/mudler/LocalAI/pkg/utils"
-)
-
-func SoundGeneration(
-	backend string,
-	modelFile string,
-	text string,
-	duration *float32,
-	temperature *float32,
-	doSample *bool,
-	sourceFile *string,
-	sourceDivisor *int32,
-	loader *model.ModelLoader,
-	appConfig *config.ApplicationConfig,
-	backendConfig config.BackendConfig,
-) (string, *proto.Result, error) {
-	if backend == "" {
-		return "", nil, fmt.Errorf("backend is a required parameter")
-	}
-
-	grpcOpts := GRPCModelOpts(backendConfig)
-	opts := modelOpts(config.BackendConfig{}, appConfig, []model.Option{
-		model.WithBackendString(backend),
-		model.WithModel(modelFile),
-		model.WithContext(appConfig.Context),
-		model.WithAssetDir(appConfig.AssetsDestination),
-		model.WithLoadGRPCLoadModelOpts(grpcOpts),
-	})
-
-	soundGenModel, err := loader.BackendLoader(opts...)
-	if err != nil {
-		return "", nil, err
-	}
-
-	if soundGenModel == nil {
-		return "", nil, fmt.Errorf("could not load sound generation model")
-	}
-
-	if err := os.MkdirAll(appConfig.AudioDir, 0750); err != nil {
-		return "", nil, fmt.Errorf("failed creating audio directory: %s", err)
-	}
-
-	fileName := utils.GenerateUniqueFileName(appConfig.AudioDir, "sound_generation", ".wav")
-	filePath := filepath.Join(appConfig.AudioDir, fileName)
-
-	res, err := soundGenModel.SoundGeneration(context.Background(), &proto.SoundGenerationRequest{
-		Text:        text,
-		Model:       modelFile,
-		Dst:         filePath,
-		Sample:      doSample,
-		Duration:    duration,
-		Temperature: temperature,
-		Src:         sourceFile,
-		SrcDivisor:  sourceDivisor,
-	})
-
-	// return RPC error if any
-	if !res.Success {
-		return "", nil, fmt.Errorf(res.Message)
-	}
-
-	return filePath, res, err
-}
--- a/core/backend/transcript.go
+++ b/core/backend/transcript.go
@@ -3,13 +3,12 @@ package backend
 import (
 	"context"
 	"fmt"
-	"time"

 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/schema"

 	"github.com/mudler/LocalAI/pkg/grpc/proto"
-	"github.com/mudler/LocalAI/pkg/model"
+	model "github.com/mudler/LocalAI/pkg/model"
 )

 func ModelTranscription(audio, language string, translate bool, ml *model.ModelLoader, backendConfig config.BackendConfig, appConfig *config.ApplicationConfig) (*schema.TranscriptionResult, error) {
@@ -22,40 +21,19 @@ func ModelTranscription(audio, language string, translate bool, ml *model.ModelL
 		model.WithAssetDir(appConfig.AssetsDestination),
 	})

-	transcriptionModel, err := ml.BackendLoader(opts...)
+	whisperModel, err := ml.BackendLoader(opts...)
 	if err != nil {
 		return nil, err
 	}

-	if transcriptionModel == nil {
-		return nil, fmt.Errorf("could not load transcription model")
+	if whisperModel == nil {
+		return nil, fmt.Errorf("could not load whisper model")
 	}

-	r, err := transcriptionModel.AudioTranscription(context.Background(), &proto.TranscriptRequest{
+	return whisperModel.AudioTranscription(context.Background(), &proto.TranscriptRequest{
 		Dst:       audio,
 		Language:  language,
 		Translate: translate,
 		Threads:   uint32(*backendConfig.Threads),
 	})
-	if err != nil {
-		return nil, err
-	}
-	tr := &schema.TranscriptionResult{
-		Text: r.Text,
-	}
-	for _, s := range r.Segments {
-		var tks []int
-		for _, t := range s.Tokens {
-			tks = append(tks, int(t))
-		}
-		tr.Segments = append(tr.Segments,
-			schema.Segment{
-				Text:   s.Text,
-				Id:     int(s.Id),
-				Start:  time.Duration(s.Start),
-				End:    time.Duration(s.End),
-				Tokens: tks,
-			})
-	}
-	return tr, err
 }
--- a/core/backend/tts.go
+++ b/core/backend/tts.go
@@ -9,15 +9,31 @@ import (
 	"github.com/mudler/LocalAI/core/config"

 	"github.com/mudler/LocalAI/pkg/grpc/proto"
-	"github.com/mudler/LocalAI/pkg/model"
+	model "github.com/mudler/LocalAI/pkg/model"
 	"github.com/mudler/LocalAI/pkg/utils"
 )

+func generateUniqueFileName(dir, baseName, ext string) string {
+	counter := 1
+	fileName := baseName + ext
+
+	for {
+		filePath := filepath.Join(dir, fileName)
+		_, err := os.Stat(filePath)
+		if os.IsNotExist(err) {
+			return fileName
+		}
+
+		counter++
+		fileName = fmt.Sprintf("%s_%d%s", baseName, counter, ext)
+	}
+}
+
 func ModelTTS(
 	backend,
 	text,
 	modelFile,
-	voice,
+	voice ,
 	language string,
 	loader *model.ModelLoader,
 	appConfig *config.ApplicationConfig,
@@ -28,7 +44,7 @@ func ModelTTS(
 		bb = model.PiperBackend
 	}

-	grpcOpts := GRPCModelOpts(backendConfig)
+	grpcOpts := gRPCModelOpts(backendConfig)

 	opts := modelOpts(config.BackendConfig{}, appConfig, []model.Option{
 		model.WithBackendString(bb),
@@ -50,7 +66,7 @@ func ModelTTS(
 		return "", nil, fmt.Errorf("failed creating audio directory: %s", err)
 	}

-	fileName := utils.GenerateUniqueFileName(appConfig.AudioDir, "tts", ".wav")
+	fileName := generateUniqueFileName(appConfig.AudioDir, "tts", ".wav")
 	filePath := filepath.Join(appConfig.AudioDir, fileName)

 	// If the model file is not empty, we pass it joined with the model path
@@ -72,15 +88,12 @@ func ModelTTS(
 	}

 	res, err := ttsModel.TTS(context.Background(), &proto.TTSRequest{
-		Text:     text,
-		Model:    modelPath,
-		Voice:    voice,
-		Dst:      filePath,
+		Text:  text,
+		Model: modelPath,
+		Voice: voice,
+		Dst:   filePath,
 		Language: &language,
 	})
-	if err != nil {
-		return "", nil, err
-	}

 	// return RPC error if any
 	if !res.Success {
--- a/core/cli/api/p2p.go
+++ b/core/cli/api/p2p.go
@@ -1,80 +0,0 @@
-package cli_api
-
-import (
-	"context"
-	"fmt"
-	"net"
-	"os"
-	"strings"
-
-	"github.com/mudler/LocalAI/core/p2p"
-	"github.com/mudler/edgevpn/pkg/node"
-
-	"github.com/rs/zerolog/log"
-)
-
-func StartP2PStack(ctx context.Context, address, token, networkID string, federated bool) error {
-	var n *node.Node
-	// Here we are avoiding creating multiple nodes:
-	// - if the federated mode is enabled, we create a federated node and expose a service
-	// - exposing a service creates a node with specific options, and we don't want to create another node
-
-	// If the federated mode is enabled, we expose a service to the local instance running
-	// at r.Address
-	if federated {
-		_, port, err := net.SplitHostPort(address)
-		if err != nil {
-			return err
-		}
-
-		// Here a new node is created and started
-		// and a service is exposed by the node
-		node, err := p2p.ExposeService(ctx, "localhost", port, token, p2p.NetworkID(networkID, p2p.FederatedID))
-		if err != nil {
-			return err
-		}
-
-		if err := p2p.ServiceDiscoverer(ctx, node, token, p2p.NetworkID(networkID, p2p.FederatedID), nil, false); err != nil {
-			return err
-		}
-
-		n = node
-	}
-
-	// If the p2p mode is enabled, we start the service discovery
-	if token != "" {
-		// If a node wasn't created previously, create it
-		if n == nil {
-			node, err := p2p.NewNode(token)
-			if err != nil {
-				return err
-			}
-			err = node.Start(ctx)
-			if err != nil {
-				return fmt.Errorf("starting new node: %w", err)
-			}
-			n = node
-		}
-
-		// Attach a ServiceDiscoverer to the p2p node
-		log.Info().Msg("Starting P2P server discovery...")
-		if err := p2p.ServiceDiscoverer(ctx, n, token, p2p.NetworkID(networkID, p2p.WorkerID), func(serviceID string, node p2p.NodeData) {
-			var tunnelAddresses []string
-			for _, v := range p2p.GetAvailableNodes(p2p.NetworkID(networkID, p2p.WorkerID)) {
-				if v.IsOnline() {
-					tunnelAddresses = append(tunnelAddresses, v.TunnelAddress)
-				} else {
-					log.Info().Msgf("Node %s is offline", v.ID)
-				}
-			}
-			tunnelEnvVar := strings.Join(tunnelAddresses, ",")
-
-			os.Setenv("LLAMACPP_GRPC_SERVERS", tunnelEnvVar)
-			log.Debug().Msgf("setting LLAMACPP_GRPC_SERVERS to %s", tunnelEnvVar)
-		}, true); err != nil {
-			return err
-		}
-	}
-
-	return nil
-}
--- a/core/cli/cli.go
+++ b/core/cli/cli.go
@@ -8,13 +8,12 @@ import (
 var CLI struct {
 	cliContext.Context `embed:""`

-	Run             RunCMD             `cmd:"" help:"Run LocalAI, this the default command if no other command is specified. Run 'local-ai run --help' for more information" default:"withargs"`
-	Federated       FederatedCLI       `cmd:"" help:"Run LocalAI in federated mode"`
-	Models          ModelsCMD          `cmd:"" help:"Manage LocalAI models and definitions"`
-	TTS             TTSCMD             `cmd:"" help:"Convert text to speech"`
-	SoundGeneration SoundGenerationCMD `cmd:"" help:"Generates audio files from text or audio"`
-	Transcript      TranscriptCMD      `cmd:"" help:"Convert audio to text"`
-	Worker          worker.Worker      `cmd:"" help:"Run workers to distribute workload (llama.cpp-only)"`
-	Util            UtilCMD            `cmd:"" help:"Utility commands"`
-	Explorer        ExplorerCMD        `cmd:"" help:"Run p2p explorer"`
+	Run        RunCMD        `cmd:"" help:"Run LocalAI, this the default command if no other command is specified. Run 'local-ai run --help' for more information" default:"withargs"`
+	Federated  FederatedCLI  `cmd:"" help:"Run LocalAI in federated mode"`
+	Models     ModelsCMD     `cmd:"" help:"Manage LocalAI models and definitions"`
+	TTS        TTSCMD        `cmd:"" help:"Convert text to speech"`
+	Transcript TranscriptCMD `cmd:"" help:"Convert audio to text"`
+	Worker     worker.Worker `cmd:"" help:"Run workers to distribute workload (llama.cpp-only)"`
+	Util       UtilCMD       `cmd:"" help:"Utility commands"`
+	Explorer   ExplorerCMD   `cmd:"" help:"Run p2p explorer"`
 }
--- a/core/cli/run.go
+++ b/core/cli/run.go
@@ -3,10 +3,11 @@ package cli
 import (
 	"context"
 	"fmt"
+	"net"
+	"os"
 	"strings"
 	"time"

-	cli_api "github.com/mudler/LocalAI/core/cli/api"
 	cliContext "github.com/mudler/LocalAI/core/cli/context"
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/http"
@@ -41,35 +42,29 @@ type RunCMD struct {
 	Threads     int  `env:"LOCALAI_THREADS,THREADS" short:"t" help:"Number of threads used for parallel computation. Usage of the number of physical cores in the system is suggested" group:"performance"`
 	ContextSize int  `env:"LOCALAI_CONTEXT_SIZE,CONTEXT_SIZE" default:"512" help:"Default context size for models" group:"performance"`

-	Address                            string   `env:"LOCALAI_ADDRESS,ADDRESS" default:":8080" help:"Bind address for the API server" group:"api"`
-	CORS                               bool     `env:"LOCALAI_CORS,CORS" help:"" group:"api"`
-	CORSAllowOrigins                   string   `env:"LOCALAI_CORS_ALLOW_ORIGINS,CORS_ALLOW_ORIGINS" group:"api"`
-	LibraryPath                        string   `env:"LOCALAI_LIBRARY_PATH,LIBRARY_PATH" help:"Path to the library directory (for e.g. external libraries used by backends)" default:"/usr/share/local-ai/libs" group:"backends"`
-	CSRF                               bool     `env:"LOCALAI_CSRF" help:"Enables fiber CSRF middleware" group:"api"`
-	UploadLimit                        int      `env:"LOCALAI_UPLOAD_LIMIT,UPLOAD_LIMIT" default:"15" help:"Default upload-limit in MB" group:"api"`
-	APIKeys                            []string `env:"LOCALAI_API_KEY,API_KEY" help:"List of API Keys to enable API authentication. When this is set, all the requests must be authenticated with one of these API keys" group:"api"`
-	DisableWebUI                       bool     `env:"LOCALAI_DISABLE_WEBUI,DISABLE_WEBUI" default:"false" help:"Disable webui" group:"api"`
-	DisablePredownloadScan             bool     `env:"LOCALAI_DISABLE_PREDOWNLOAD_SCAN" help:"If true, disables the best-effort security scanner before downloading any files." group:"hardening" default:"false"`
-	OpaqueErrors                       bool     `env:"LOCALAI_OPAQUE_ERRORS" default:"false" help:"If true, all error responses are replaced with blank 500 errors. This is intended only for hardening against information leaks and is normally not recommended." group:"hardening"`
-	UseSubtleKeyComparison             bool     `env:"LOCALAI_SUBTLE_KEY_COMPARISON" default:"false" help:"If true, API Key validation comparisons will be performed using constant-time comparisons rather than simple equality. This trades off performance on each request for resiliancy against timing attacks." group:"hardening"`
-	DisableApiKeyRequirementForHttpGet bool     `env:"LOCALAI_DISABLE_API_KEY_REQUIREMENT_FOR_HTTP_GET" default:"false" help:"If true, a valid API key is not required to issue GET requests to portions of the web ui. This should only be enabled in secure testing environments" group:"hardening"`
-	HttpGetExemptedEndpoints           []string `env:"LOCALAI_HTTP_GET_EXEMPTED_ENDPOINTS" default:"^/$,^/browse/?$,^/talk/?$,^/p2p/?$,^/chat/?$,^/text2image/?$,^/tts/?$,^/static/.*$,^/swagger.*$" help:"If LOCALAI_DISABLE_API_KEY_REQUIREMENT_FOR_HTTP_GET is overriden to true, this is the list of endpoints to exempt. Only adjust this in case of a security incident or as a result of a personal security posture review" group:"hardening"`
-	Peer2Peer                          bool     `env:"LOCALAI_P2P,P2P" name:"p2p" default:"false" help:"Enable P2P mode" group:"p2p"`
-	Peer2PeerDHTInterval               int      `env:"LOCALAI_P2P_DHT_INTERVAL,P2P_DHT_INTERVAL" default:"360" name:"p2p-dht-interval" help:"Interval for DHT refresh (used during token generation)" group:"p2p"`
-	Peer2PeerOTPInterval               int      `env:"LOCALAI_P2P_OTP_INTERVAL,P2P_OTP_INTERVAL" default:"9000" name:"p2p-otp-interval" help:"Interval for OTP refresh (used during token generation)" group:"p2p"`
-	Peer2PeerToken                     string   `env:"LOCALAI_P2P_TOKEN,P2P_TOKEN,TOKEN" name:"p2ptoken" help:"Token for P2P mode (optional)" group:"p2p"`
-	Peer2PeerNetworkID                 string   `env:"LOCALAI_P2P_NETWORK_ID,P2P_NETWORK_ID" help:"Network ID for P2P mode, can be set arbitrarly by the user for grouping a set of instances" group:"p2p"`
-	ParallelRequests                   bool     `env:"LOCALAI_PARALLEL_REQUESTS,PARALLEL_REQUESTS" help:"Enable backends to handle multiple requests in parallel if they support it (e.g.: llama.cpp or vllm)" group:"backends"`
-	SingleActiveBackend                bool     `env:"LOCALAI_SINGLE_ACTIVE_BACKEND,SINGLE_ACTIVE_BACKEND" help:"Allow only one backend to be run at a time" group:"backends"`
-	PreloadBackendOnly                 bool     `env:"LOCALAI_PRELOAD_BACKEND_ONLY,PRELOAD_BACKEND_ONLY" default:"false" help:"Do not launch the API services, only the preloaded models / backends are started (useful for multi-node setups)" group:"backends"`
-	ExternalGRPCBackends               []string `env:"LOCALAI_EXTERNAL_GRPC_BACKENDS,EXTERNAL_GRPC_BACKENDS" help:"A list of external grpc backends" group:"backends"`
-	EnableWatchdogIdle                 bool     `env:"LOCALAI_WATCHDOG_IDLE,WATCHDOG_IDLE" default:"false" help:"Enable watchdog for stopping backends that are idle longer than the watchdog-idle-timeout" group:"backends"`
-	WatchdogIdleTimeout                string   `env:"LOCALAI_WATCHDOG_IDLE_TIMEOUT,WATCHDOG_IDLE_TIMEOUT" default:"15m" help:"Threshold beyond which an idle backend should be stopped" group:"backends"`
-	EnableWatchdogBusy                 bool     `env:"LOCALAI_WATCHDOG_BUSY,WATCHDOG_BUSY" default:"false" help:"Enable watchdog for stopping backends that are busy longer than the watchdog-busy-timeout" group:"backends"`
-	WatchdogBusyTimeout                string   `env:"LOCALAI_WATCHDOG_BUSY_TIMEOUT,WATCHDOG_BUSY_TIMEOUT" default:"5m" help:"Threshold beyond which a busy backend should be stopped" group:"backends"`
-	Federated                          bool     `env:"LOCALAI_FEDERATED,FEDERATED" help:"Enable federated instance" group:"federated"`
-	DisableGalleryEndpoint             bool     `env:"LOCALAI_DISABLE_GALLERY_ENDPOINT,DISABLE_GALLERY_ENDPOINT" help:"Disable the gallery endpoints" group:"api"`
-	LoadToMemory                       []string `env:"LOCALAI_LOAD_TO_MEMORY,LOAD_TO_MEMORY" help:"A list of models to load into memory at startup" group:"models"`
+	Address                string   `env:"LOCALAI_ADDRESS,ADDRESS" default:":8080" help:"Bind address for the API server" group:"api"`
+	CORS                   bool     `env:"LOCALAI_CORS,CORS" help:"" group:"api"`
+	CORSAllowOrigins       string   `env:"LOCALAI_CORS_ALLOW_ORIGINS,CORS_ALLOW_ORIGINS" group:"api"`
+	LibraryPath            string   `env:"LOCALAI_LIBRARY_PATH,LIBRARY_PATH" help:"Path to the library directory (for e.g. external libraries used by backends)" default:"/usr/share/local-ai/libs" group:"backends"`
+	CSRF                   bool     `env:"LOCALAI_CSRF" help:"Enables fiber CSRF middleware" group:"api"`
+	UploadLimit            int      `env:"LOCALAI_UPLOAD_LIMIT,UPLOAD_LIMIT" default:"15" help:"Default upload-limit in MB" group:"api"`
+	APIKeys                []string `env:"LOCALAI_API_KEY,API_KEY" help:"List of API Keys to enable API authentication. When this is set, all the requests must be authenticated with one of these API keys" group:"api"`
+	DisableWebUI           bool     `env:"LOCALAI_DISABLE_WEBUI,DISABLE_WEBUI" default:"false" help:"Disable webui" group:"api"`
+	DisablePredownloadScan bool     `env:"LOCALAI_DISABLE_PREDOWNLOAD_SCAN" help:"If true, disables the best-effort security scanner before downloading any files." group:"hardening" default:"false"`
+	OpaqueErrors           bool     `env:"LOCALAI_OPAQUE_ERRORS" default:"false" help:"If true, all error responses are replaced with blank 500 errors. This is intended only for hardening against information leaks and is normally not recommended." group:"hardening"`
+	Peer2Peer              bool     `env:"LOCALAI_P2P,P2P" name:"p2p" default:"false" help:"Enable P2P mode" group:"p2p"`
+	Peer2PeerToken         string   `env:"LOCALAI_P2P_TOKEN,P2P_TOKEN,TOKEN" name:"p2ptoken" help:"Token for P2P mode (optional)" group:"p2p"`
+	Peer2PeerNetworkID     string   `env:"LOCALAI_P2P_NETWORK_ID,P2P_NETWORK_ID" help:"Network ID for P2P mode, can be set arbitrarly by the user for grouping a set of instances" group:"p2p"`
+	ParallelRequests       bool     `env:"LOCALAI_PARALLEL_REQUESTS,PARALLEL_REQUESTS" help:"Enable backends to handle multiple requests in parallel if they support it (e.g.: llama.cpp or vllm)" group:"backends"`
+	SingleActiveBackend    bool     `env:"LOCALAI_SINGLE_ACTIVE_BACKEND,SINGLE_ACTIVE_BACKEND" help:"Allow only one backend to be run at a time" group:"backends"`
+	PreloadBackendOnly     bool     `env:"LOCALAI_PRELOAD_BACKEND_ONLY,PRELOAD_BACKEND_ONLY" default:"false" help:"Do not launch the API services, only the preloaded models / backends are started (useful for multi-node setups)" group:"backends"`
+	ExternalGRPCBackends   []string `env:"LOCALAI_EXTERNAL_GRPC_BACKENDS,EXTERNAL_GRPC_BACKENDS" help:"A list of external grpc backends" group:"backends"`
+	EnableWatchdogIdle     bool     `env:"LOCALAI_WATCHDOG_IDLE,WATCHDOG_IDLE" default:"false" help:"Enable watchdog for stopping backends that are idle longer than the watchdog-idle-timeout" group:"backends"`
+	WatchdogIdleTimeout    string   `env:"LOCALAI_WATCHDOG_IDLE_TIMEOUT,WATCHDOG_IDLE_TIMEOUT" default:"15m" help:"Threshold beyond which an idle backend should be stopped" group:"backends"`
+	EnableWatchdogBusy     bool     `env:"LOCALAI_WATCHDOG_BUSY,WATCHDOG_BUSY" default:"false" help:"Enable watchdog for stopping backends that are busy longer than the watchdog-busy-timeout" group:"backends"`
+	WatchdogBusyTimeout    string   `env:"LOCALAI_WATCHDOG_BUSY_TIMEOUT,WATCHDOG_BUSY_TIMEOUT" default:"5m" help:"Threshold beyond which a busy backend should be stopped" group:"backends"`
+	Federated              bool     `env:"LOCALAI_FEDERATED,FEDERATED" help:"Enable federated instance" group:"federated"`
+	DisableGalleryEndpoint bool     `env:"LOCALAI_DISABLE_GALLERY_ENDPOINT,DISABLE_GALLERY_ENDPOINT" help:"Disable the gallery endpoints" group:"api"`
 }

 func (r *RunCMD) Run(ctx *cliContext.Context) error {
@@ -101,11 +96,7 @@ func (r *RunCMD) Run(ctx *cliContext.Context) error {
 		config.WithModelsURL(append(r.Models, r.ModelArgs...)...),
 		config.WithOpaqueErrors(r.OpaqueErrors),
 		config.WithEnforcedPredownloadScans(!r.DisablePredownloadScan),
-		config.WithSubtleKeyComparison(r.UseSubtleKeyComparison),
-		config.WithDisableApiKeyRequirementForHttpGet(r.DisableApiKeyRequirementForHttpGet),
-		config.WithHttpGetExemptedEndpoints(r.HttpGetExemptedEndpoints),
 		config.WithP2PNetworkID(r.Peer2PeerNetworkID),
-		config.WithLoadToMemory(r.LoadToMemory),
 	}

 	token := ""
@@ -116,7 +107,7 @@ func (r *RunCMD) Run(ctx *cliContext.Context) error {
 			// IF no token is provided, and p2p is enabled,
 			// we generate one and wait for the user to pick up the token (this is for interactive)
 			log.Info().Msg("No token provided, generating one")
-			token = p2p.GenerateToken(r.Peer2PeerDHTInterval, r.Peer2PeerOTPInterval)
+			token = p2p.GenerateToken()
 			log.Info().Msg("Generated Token:")
 			fmt.Println(token)

@@ -124,12 +115,52 @@ func (r *RunCMD) Run(ctx *cliContext.Context) error {
 			fmt.Printf("export TOKEN=\"%s\"\nlocal-ai worker p2p-llama-cpp-rpc\n", token)
 		}
 		opts = append(opts, config.WithP2PToken(token))
+
+		node, err := p2p.NewNode(token)
+		if err != nil {
+			return err
+		}
+		nodeContext := context.Background()
+
+		err = node.Start(nodeContext)
+		if err != nil {
+			return fmt.Errorf("starting new node: %w", err)
+		}
+
+		log.Info().Msg("Starting P2P server discovery...")
+		if err := p2p.ServiceDiscoverer(nodeContext, node, token, p2p.NetworkID(r.Peer2PeerNetworkID, p2p.WorkerID), func(serviceID string, node p2p.NodeData) {
+			var tunnelAddresses []string
+			for _, v := range p2p.GetAvailableNodes(p2p.NetworkID(r.Peer2PeerNetworkID, p2p.WorkerID)) {
+				if v.IsOnline() {
+					tunnelAddresses = append(tunnelAddresses, v.TunnelAddress)
+				} else {
+					log.Info().Msgf("Node %s is offline", v.ID)
+				}
+			}
+			tunnelEnvVar := strings.Join(tunnelAddresses, ",")
+
+			os.Setenv("LLAMACPP_GRPC_SERVERS", tunnelEnvVar)
+			log.Debug().Msgf("setting LLAMACPP_GRPC_SERVERS to %s", tunnelEnvVar)
+		}, true); err != nil {
+			return err
+		}
 	}

-	backgroundCtx := context.Background()
+	if r.Federated {
+		_, port, err := net.SplitHostPort(r.Address)
+		if err != nil {
+			return err
+		}
+		fedCtx := context.Background()

-	if err := cli_api.StartP2PStack(backgroundCtx, r.Address, token, r.Peer2PeerNetworkID, r.Federated); err != nil {
-		return err
+		node, err := p2p.ExposeService(fedCtx, "localhost", port, token, p2p.NetworkID(r.Peer2PeerNetworkID, p2p.FederatedID))
+		if err != nil {
+			return err
+		}
+
+		if err := p2p.ServiceDiscoverer(fedCtx, node, token, p2p.NetworkID(r.Peer2PeerNetworkID, p2p.FederatedID), nil, false); err != nil {
+			return err
+		}
 	}

 	idleWatchDog := r.EnableWatchdogIdle
--- a/core/cli/soundgeneration.go
+++ b/core/cli/soundgeneration.go
@@ -1,110 +0,0 @@
-package cli
-
-import (
-	"context"
-	"fmt"
-	"os"
-	"path/filepath"
-	"strconv"
-	"strings"
-
-	"github.com/mudler/LocalAI/core/backend"
-	cliContext "github.com/mudler/LocalAI/core/cli/context"
-	"github.com/mudler/LocalAI/core/config"
-	"github.com/mudler/LocalAI/pkg/model"
-	"github.com/rs/zerolog/log"
-)
-
-type SoundGenerationCMD struct {
-	Text []string `arg:""`
-
-	Backend                string   `short:"b" required:"" help:"Backend to run the SoundGeneration model"`
-	Model                  string   `short:"m" required:"" help:"Model name to run the SoundGeneration"`
-	Duration               string   `short:"d" help:"If specified, the length of audio to generate in seconds"`
-	Temperature            string   `short:"t" help:"If specified, the temperature of the generation"`
-	InputFile              string   `short:"i" help:"If specified, the input file to condition generation upon"`
-	InputFileSampleDivisor string   `short:"f" help:"If InputFile and this divisor is specified, the first portion of the sample file will be used"`
-	DoSample               bool     `short:"s" default:"true" help:"Enables sampling from the model. Better quality at the cost of speed. Defaults to enabled."`
-	OutputFile             string   `short:"o" type:"path" help:"The path to write the output wav file"`
-	ModelsPath             string   `env:"LOCALAI_MODELS_PATH,MODELS_PATH" type:"path" default:"${basepath}/models" help:"Path containing models used for inferencing" group:"storage"`
-	BackendAssetsPath      string   `env:"LOCALAI_BACKEND_ASSETS_PATH,BACKEND_ASSETS_PATH" type:"path" default:"/tmp/localai/backend_data" help:"Path used to extract libraries that are required by some of the backends in runtime" group:"storage"`
-	ExternalGRPCBackends   []string `env:"LOCALAI_EXTERNAL_GRPC_BACKENDS,EXTERNAL_GRPC_BACKENDS" help:"A list of external grpc backends" group:"backends"`
-}
-
-func parseToFloat32Ptr(input string) *float32 {
-	f, err := strconv.ParseFloat(input, 32)
-	if err != nil {
-		return nil
-	}
-	f2 := float32(f)
-	return &f2
-}
-
-func parseToInt32Ptr(input string) *int32 {
-	i, err := strconv.ParseInt(input, 10, 32)
-	if err != nil {
-		return nil
-	}
-	i2 := int32(i)
-	return &i2
-}
-
-func (t *SoundGenerationCMD) Run(ctx *cliContext.Context) error {
-	outputFile := t.OutputFile
-	outputDir := t.BackendAssetsPath
-	if outputFile != "" {
-		outputDir = filepath.Dir(outputFile)
-	}
-
-	text := strings.Join(t.Text, " ")
-
-	externalBackends := make(map[string]string)
-	// split ":" to get backend name and the uri
-	for _, v := range t.ExternalGRPCBackends {
-		backend := v[:strings.IndexByte(v, ':')]
-		uri := v[strings.IndexByte(v, ':')+1:]
-		externalBackends[backend] = uri
-		fmt.Printf("TMP externalBackends[%q]=%q\n\n", backend, uri)
-	}
-
-	opts := &config.ApplicationConfig{
-		ModelPath:            t.ModelsPath,
-		Context:              context.Background(),
-		AudioDir:             outputDir,
-		AssetsDestination:    t.BackendAssetsPath,
-		ExternalGRPCBackends: externalBackends,
-	}
-	ml := model.NewModelLoader(opts.ModelPath)
-
-	defer func() {
-		err := ml.StopAllGRPC()
-		if err != nil {
-			log.Error().Err(err).Msg("unable to stop all grpc processes")
-		}
-	}()
-
-	options := config.BackendConfig{}
-	options.SetDefaults()
-
-	var inputFile *string
-	if t.InputFile != "" {
-		inputFile = &t.InputFile
-	}
-
-	filePath, _, err := backend.SoundGeneration(t.Backend, t.Model, text,
-		parseToFloat32Ptr(t.Duration), parseToFloat32Ptr(t.Temperature), &t.DoSample,
-		inputFile, parseToInt32Ptr(t.InputFileSampleDivisor), ml, opts, options)
-
-	if err != nil {
-		return err
-	}
-	if outputFile != "" {
-		if err := os.Rename(filePath, outputFile); err != nil {
-			return err
-		}
-		fmt.Printf("Generate file %s\n", outputFile)
-	} else {
-		fmt.Printf("Generate file %s\n", filePath)
-	}
-	return nil
-}
--- a/core/cli/worker/worker.go
+++ b/core/cli/worker/worker.go
@@ -2,7 +2,6 @@ package worker

 type WorkerFlags struct {
 	BackendAssetsPath string `env:"LOCALAI_BACKEND_ASSETS_PATH,BACKEND_ASSETS_PATH" type:"path" default:"/tmp/localai/backend_data" help:"Path used to extract libraries that are required by some of the backends in runtime" group:"storage"`
-	ExtraLLamaCPPArgs string `name:"llama-cpp-args" env:"LOCALAI_EXTRA_LLAMA_CPP_ARGS,EXTRA_LLAMA_CPP_ARGS" help:"Extra arguments to pass to llama-cpp-rpc-server"`
 }

 type Worker struct {
--- a/core/cli/worker/worker_llamacpp.go
+++ b/core/cli/worker/worker_llamacpp.go
@@ -3,7 +3,6 @@ package worker
 import (
 	"fmt"
 	"os"
-	"strings"
 	"syscall"

 	cliContext "github.com/mudler/LocalAI/core/cli/context"
@@ -13,6 +12,7 @@ import (
 )

 type LLamaCPP struct {
+	Args        []string `arg:"" optional:"" name:"models" help:"Model configuration URLs to load"`
 	WorkerFlags `embed:""`
 }

@@ -34,8 +34,9 @@ func (r *LLamaCPP) Run(ctx *cliContext.Context) error {
 		"llama-cpp-rpc-server",
 	)

-	args := strings.Split(r.ExtraLLamaCPPArgs, " ")
+	args := os.Args[4:]
 	args, grpcProcess = library.LoadLDSO(r.BackendAssetsPath, args, grpcProcess)
+
 	args = append([]string{grpcProcess}, args...)
 	return syscall.Exec(
 		grpcProcess,
--- a/core/cli/worker/worker_p2p.go
+++ b/core/cli/worker/worker_p2p.go
@@ -8,7 +8,6 @@ import (
 	"fmt"
 	"os"
 	"os/exec"
-	"strings"
 	"time"

 	cliContext "github.com/mudler/LocalAI/core/cli/context"
@@ -21,11 +20,12 @@ import (

 type P2P struct {
 	WorkerFlags        `embed:""`
-	Token              string `env:"LOCALAI_TOKEN,LOCALAI_P2P_TOKEN,TOKEN" help:"P2P token to use"`
-	NoRunner           bool   `env:"LOCALAI_NO_RUNNER,NO_RUNNER" help:"Do not start the llama-cpp-rpc-server"`
-	RunnerAddress      string `env:"LOCALAI_RUNNER_ADDRESS,RUNNER_ADDRESS" help:"Address of the llama-cpp-rpc-server"`
-	RunnerPort         string `env:"LOCALAI_RUNNER_PORT,RUNNER_PORT" help:"Port of the llama-cpp-rpc-server"`
-	Peer2PeerNetworkID string `env:"LOCALAI_P2P_NETWORK_ID,P2P_NETWORK_ID" help:"Network ID for P2P mode, can be set arbitrarly by the user for grouping a set of instances" group:"p2p"`
+	Token              string   `env:"LOCALAI_TOKEN,LOCALAI_P2P_TOKEN,TOKEN" help:"P2P token to use"`
+	NoRunner           bool     `env:"LOCALAI_NO_RUNNER,NO_RUNNER" help:"Do not start the llama-cpp-rpc-server"`
+	RunnerAddress      string   `env:"LOCALAI_RUNNER_ADDRESS,RUNNER_ADDRESS" help:"Address of the llama-cpp-rpc-server"`
+	RunnerPort         string   `env:"LOCALAI_RUNNER_PORT,RUNNER_PORT" help:"Port of the llama-cpp-rpc-server"`
+	ExtraLLamaCPPArgs  []string `env:"LOCALAI_EXTRA_LLAMA_CPP_ARGS,EXTRA_LLAMA_CPP_ARGS" help:"Extra arguments to pass to llama-cpp-rpc-server"`
+	Peer2PeerNetworkID string   `env:"LOCALAI_P2P_NETWORK_ID,P2P_NETWORK_ID" help:"Network ID for P2P mode, can be set arbitrarly by the user for grouping a set of instances" group:"p2p"`
 }

 func (r *P2P) Run(ctx *cliContext.Context) error {
@@ -76,8 +76,8 @@ func (r *P2P) Run(ctx *cliContext.Context) error {
 					"util",
 					"llama-cpp-rpc-server",
 				)
-				extraArgs := strings.Split(r.ExtraLLamaCPPArgs, " ")
-				args := append([]string{"--host", address, "--port", fmt.Sprint(port)}, extraArgs...)
+
+				args := append([]string{"--host", address, "--port", fmt.Sprint(port)}, r.ExtraLLamaCPPArgs...)
 				args, grpcProcess = library.LoadLDSO(r.BackendAssetsPath, args, grpcProcess)

 				cmd := exec.Command(
--- a/core/config/application_config.go
+++ b/core/config/application_config.go
@@ -4,7 +4,6 @@ import (
 	"context"
 	"embed"
 	"encoding/json"
-	"regexp"
 	"time"

 	"github.com/mudler/LocalAI/pkg/xsysinfo"
@@ -17,6 +16,7 @@ type ApplicationConfig struct {
 	ModelPath                           string
 	LibPath                             string
 	UploadLimitMB, Threads, ContextSize int
+	DisableWebUI                        bool
 	F16                                 bool
 	Debug                               bool
 	ImageDir                            string
@@ -31,18 +31,11 @@ type ApplicationConfig struct {
 	PreloadModelsFromPath               string
 	CORSAllowOrigins                    string
 	ApiKeys                             []string
+	EnforcePredownloadScans             bool
+	OpaqueErrors                        bool
 	P2PToken                            string
 	P2PNetworkID                        string

-	DisableWebUI                       bool
-	EnforcePredownloadScans            bool
-	OpaqueErrors                       bool
-	UseSubtleKeyComparison             bool
-	DisableApiKeyRequirementForHttpGet bool
-	HttpGetExemptedEndpoints           []*regexp.Regexp
-	DisableGalleryEndpoint             bool
-	LoadToMemory                       []string
-
 	ModelLibraryURL string

 	Galleries []Gallery
@@ -64,6 +57,8 @@ type ApplicationConfig struct {
 	ModelsURL []string

 	WatchDogBusyTimeout, WatchDogIdleTimeout time.Duration
+
+	DisableGalleryEndpoint bool
 }

 type AppOption func(*ApplicationConfig)
@@ -332,38 +327,6 @@ func WithOpaqueErrors(opaque bool) AppOption {
 	}
 }

-func WithLoadToMemory(models []string) AppOption {
-	return func(o *ApplicationConfig) {
-		o.LoadToMemory = models
-	}
-}
-
-func WithSubtleKeyComparison(subtle bool) AppOption {
-	return func(o *ApplicationConfig) {
-		o.UseSubtleKeyComparison = subtle
-	}
-}
-
-func WithDisableApiKeyRequirementForHttpGet(required bool) AppOption {
-	return func(o *ApplicationConfig) {
-		o.DisableApiKeyRequirementForHttpGet = required
-	}
-}
-
-func WithHttpGetExemptedEndpoints(endpoints []string) AppOption {
-	return func(o *ApplicationConfig) {
-		o.HttpGetExemptedEndpoints = []*regexp.Regexp{}
-		for _, epr := range endpoints {
-			r, err := regexp.Compile(epr)
-			if err == nil && r != nil {
-				o.HttpGetExemptedEndpoints = append(o.HttpGetExemptedEndpoints, r)
-			} else {
-				log.Warn().Err(err).Str("regex", epr).Msg("Error while compiling HTTP Get Exemption regex, skipping this entry.")
-			}
-		}
-	}
-}
-
 // ToConfigLoaderOptions returns a slice of ConfigLoader Option.
 // Some options defined at the application level are going to be passed as defaults for
 // all the configuration for the models.
--- a/core/config/backend_config.go
+++ b/core/config/backend_config.go
@@ -126,7 +126,6 @@ type LLMConfig struct {
 	Grammar         string   `yaml:"grammar"`
 	StopWords       []string `yaml:"stopwords"`
 	Cutstrings      []string `yaml:"cutstrings"`
-	ExtractRegex    []string `yaml:"extract_regex"`
 	TrimSpace       []string `yaml:"trimspace"`
 	TrimSuffix      []string `yaml:"trimsuffix"`

--- a/core/gallery/gallery.go
+++ b/core/gallery/gallery.go
@@ -132,7 +132,7 @@ func AvailableGalleryModels(galleries []config.Gallery, basePath string) ([]*Gal
 func findGalleryURLFromReferenceURL(url string, basePath string) (string, error) {
 	var refFile string
 	uri := downloader.URI(url)
-	err := uri.DownloadWithCallback(basePath, func(url string, d []byte) error {
+	err := uri.DownloadAndUnmarshal(basePath, func(url string, d []byte) error {
 		refFile = string(d)
 		if len(refFile) == 0 {
 			return fmt.Errorf("invalid reference file at url %s: %s", url, d)
@@ -156,7 +156,7 @@ func getGalleryModels(gallery config.Gallery, basePath string) ([]*GalleryModel,
 	}
 	uri := downloader.URI(gallery.URL)

-	err := uri.DownloadWithCallback(basePath, func(url string, d []byte) error {
+	err := uri.DownloadAndUnmarshal(basePath, func(url string, d []byte) error {
 		return yaml.Unmarshal(d, &models)
 	})
 	if err != nil {
--- a/core/gallery/models.go
+++ b/core/gallery/models.go
@@ -69,7 +69,7 @@ type PromptTemplate struct {
 func GetGalleryConfigFromURL(url string, basePath string) (Config, error) {
 	var config Config
 	uri := downloader.URI(url)
-	err := uri.DownloadWithCallback(basePath, func(url string, d []byte) error {
+	err := uri.DownloadAndUnmarshal(basePath, func(url string, d []byte) error {
 		return yaml.Unmarshal(d, &config)
 	})
 	if err != nil {
--- a/core/http/app.go
+++ b/core/http/app.go
@@ -3,15 +3,13 @@ package http
 import (
 	"embed"
 	"errors"
-	"fmt"
 	"net/http"
+	"strings"

-	"github.com/dave-gray101/v2keyauth"
 	"github.com/mudler/LocalAI/pkg/utils"

 	"github.com/mudler/LocalAI/core/http/endpoints/localai"
 	"github.com/mudler/LocalAI/core/http/endpoints/openai"
-	"github.com/mudler/LocalAI/core/http/middleware"
 	"github.com/mudler/LocalAI/core/http/routes"

 	"github.com/mudler/LocalAI/core/config"
@@ -31,6 +29,24 @@ import (
 	"github.com/rs/zerolog/log"
 )

+func readAuthHeader(c *fiber.Ctx) string {
+	authHeader := c.Get("Authorization")
+
+	// elevenlabs
+	xApiKey := c.Get("xi-api-key")
+	if xApiKey != "" {
+		authHeader = "Bearer " + xApiKey
+	}
+
+	// anthropic
+	xApiKey = c.Get("x-api-key")
+	if xApiKey != "" {
+		authHeader = "Bearer " + xApiKey
+	}
+
+	return authHeader
+}
+
 // Embed a directory
 //
 //go:embed static/*
@@ -121,13 +137,36 @@ func App(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *confi
 		})
 	}

-	kaConfig, err := middleware.GetKeyAuthConfig(appConfig)
-	if err != nil || kaConfig == nil {
-		return nil, fmt.Errorf("failed to create key auth config: %w", err)
-	}
+	// Auth middleware checking if API key is valid. If no API key is set, no auth is required.
+	auth := func(c *fiber.Ctx) error {
+		if len(appConfig.ApiKeys) == 0 {
+			return c.Next()
+		}

-	// Auth is applied to _all_ endpoints. No exceptions. Filtering out endpoints to bypass is the role of the Filter property of the KeyAuth Configuration
-	app.Use(v2keyauth.New(*kaConfig))
+		if len(appConfig.ApiKeys) == 0 {
+			return c.Next()
+		}
+
+		authHeader := readAuthHeader(c)
+		if authHeader == "" {
+			return c.Status(fiber.StatusUnauthorized).JSON(fiber.Map{"message": "Authorization header missing"})
+		}
+
+		// If it's a bearer token
+		authHeaderParts := strings.Split(authHeader, " ")
+		if len(authHeaderParts) != 2 || authHeaderParts[0] != "Bearer" {
+			return c.Status(fiber.StatusUnauthorized).JSON(fiber.Map{"message": "Invalid Authorization header format"})
+		}
+
+		apiKey := authHeaderParts[1]
+		for _, key := range appConfig.ApiKeys {
+			if apiKey == key {
+				return c.Next()
+			}
+		}
+
+		return c.Status(fiber.StatusUnauthorized).JSON(fiber.Map{"message": "Invalid API key"})
+	}

 	if appConfig.CORS {
 		var c func(ctx *fiber.Ctx) error
@@ -153,13 +192,13 @@ func App(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *confi
 	galleryService := services.NewGalleryService(appConfig)
 	galleryService.Start(appConfig.Context, cl)

-	routes.RegisterElevenLabsRoutes(app, cl, ml, appConfig)
-	routes.RegisterLocalAIRoutes(app, cl, ml, appConfig, galleryService)
-	routes.RegisterOpenAIRoutes(app, cl, ml, appConfig)
+	routes.RegisterElevenLabsRoutes(app, cl, ml, appConfig, auth)
+	routes.RegisterLocalAIRoutes(app, cl, ml, appConfig, galleryService, auth)
+	routes.RegisterOpenAIRoutes(app, cl, ml, appConfig, auth)
 	if !appConfig.DisableWebUI {
-		routes.RegisterUIRoutes(app, cl, ml, appConfig, galleryService)
+		routes.RegisterUIRoutes(app, cl, ml, appConfig, galleryService, auth)
 	}
-	routes.RegisterJINARoutes(app, cl, ml, appConfig)
+	routes.RegisterJINARoutes(app, cl, ml, appConfig, auth)

 	httpFS := http.FS(embedDirStatic)

--- a/core/http/app_test.go
+++ b/core/http/app_test.go
@@ -31,9 +31,6 @@ import (
 	"github.com/sashabaranov/go-openai/jsonschema"
 )

-const apiKey = "joshua"
-const bearerKey = "Bearer " + apiKey
-
 const testPrompt = `### System:
 You are an AI assistant that follows instruction extremely well. Help as much as you can.

@@ -53,19 +50,11 @@ type modelApplyRequest struct {

 func getModelStatus(url string) (response map[string]interface{}) {
 	// Create the HTTP request
-	req, err := http.NewRequest("GET", url, nil)
-	req.Header.Set("Content-Type", "application/json")
-	req.Header.Set("Authorization", bearerKey)
+	resp, err := http.Get(url)
 	if err != nil {
 		fmt.Println("Error creating request:", err)
 		return
 	}
-	client := &http.Client{}
-	resp, err := client.Do(req)
-	if err != nil {
-		fmt.Println("Error sending request:", err)
-		return
-	}
 	defer resp.Body.Close()

 	body, err := io.ReadAll(resp.Body)
@@ -83,15 +72,14 @@ func getModelStatus(url string) (response map[string]interface{}) {
 	return
 }

-func getModels(url string) ([]gallery.GalleryModel, error) {
-	response := []gallery.GalleryModel{}
+func getModels(url string) (response []gallery.GalleryModel) {
 	uri := downloader.URI(url)
 	// TODO: No tests currently seem to exercise file:// urls. Fix?
-	err := uri.DownloadWithAuthorizationAndCallback("", bearerKey, func(url string, i []byte) error {
+	uri.DownloadAndUnmarshal("", func(url string, i []byte) error {
 		// Unmarshal YAML data into a struct
 		return json.Unmarshal(i, &response)
 	})
-	return response, err
+	return
 }

 func postModelApplyRequest(url string, request modelApplyRequest) (response map[string]interface{}) {
@@ -113,7 +101,6 @@ func postModelApplyRequest(url string, request modelApplyRequest) (response map[
 		return
 	}
 	req.Header.Set("Content-Type", "application/json")
-	req.Header.Set("Authorization", bearerKey)

 	// Make the request
 	client := &http.Client{}
@@ -153,7 +140,6 @@ func postRequestJSON[B any](url string, bodyJson *B) error {
 	}

 	req.Header.Set("Content-Type", "application/json")
-	req.Header.Set("Authorization", bearerKey)

 	client := &http.Client{}
 	resp, err := client.Do(req)
@@ -189,7 +175,6 @@ func postRequestResponseJSON[B1 any, B2 any](url string, reqJson *B1, respJson *
 	}

 	req.Header.Set("Content-Type", "application/json")
-	req.Header.Set("Authorization", bearerKey)

 	client := &http.Client{}
 	resp, err := client.Do(req)
@@ -210,35 +195,6 @@ func postRequestResponseJSON[B1 any, B2 any](url string, reqJson *B1, respJson *
 	return json.Unmarshal(body, respJson)
 }

-func postInvalidRequest(url string) (error, int) {
-
-	req, err := http.NewRequest("POST", url, bytes.NewBufferString("invalid request"))
-	if err != nil {
-		return err, -1
-	}
-
-	req.Header.Set("Content-Type", "application/json")
-
-	client := &http.Client{}
-	resp, err := client.Do(req)
-	if err != nil {
-		return err, -1
-	}
-
-	defer resp.Body.Close()
-
-	body, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return err, -1
-	}
-
-	if resp.StatusCode < 200 || resp.StatusCode >= 400 {
-		return fmt.Errorf("unexpected status code: %d, body: %s", resp.StatusCode, string(body)), resp.StatusCode
-	}
-
-	return nil, resp.StatusCode
-}
-
 //go:embed backend-assets/*
 var backendAssets embed.FS

@@ -304,7 +260,6 @@ var _ = Describe("API test", func() {
 					config.WithContext(c),
 					config.WithGalleries(galleries),
 					config.WithModelPath(modelDir),
-					config.WithApiKeys([]string{apiKey}),
 					config.WithBackendAssets(backendAssets),
 					config.WithBackendAssetsOutput(backendAssetsDir))...)
 			Expect(err).ToNot(HaveOccurred())
@@ -314,7 +269,7 @@ var _ = Describe("API test", func() {

 			go app.Listen("127.0.0.1:9090")

-			defaultConfig := openai.DefaultConfig(apiKey)
+			defaultConfig := openai.DefaultConfig("")
 			defaultConfig.BaseURL = "http://127.0.0.1:9090/v1"

 			client2 = openaigo.NewClient("")
@@ -340,19 +295,10 @@ var _ = Describe("API test", func() {
 			Expect(err).To(HaveOccurred())
 		})

-		Context("Auth Tests", func() {
-			It("Should fail if the api key is missing", func() {
-				err, sc := postInvalidRequest("http://127.0.0.1:9090/models/available")
-				Expect(err).ToNot(BeNil())
-				Expect(sc).To(Equal(403))
-			})
-		})
-
 		Context("Applying models", func() {

 			It("applies models from a gallery", func() {
-				models, err := getModels("http://127.0.0.1:9090/models/available")
-				Expect(err).To(BeNil())
+				models := getModels("http://127.0.0.1:9090/models/available")
 				Expect(len(models)).To(Equal(2), fmt.Sprint(models))
 				Expect(models[0].Installed).To(BeFalse(), fmt.Sprint(models))
 				Expect(models[1].Installed).To(BeFalse(), fmt.Sprint(models))
@@ -385,8 +331,7 @@ var _ = Describe("API test", func() {
 				Expect(content["backend"]).To(Equal("bert-embeddings"))
 				Expect(content["foo"]).To(Equal("bar"))

-				models, err = getModels("http://127.0.0.1:9090/models/available")
-				Expect(err).To(BeNil())
+				models = getModels("http://127.0.0.1:9090/models/available")
 				Expect(len(models)).To(Equal(2), fmt.Sprint(models))
 				Expect(models[0].Name).To(Or(Equal("bert"), Equal("bert2")))
 				Expect(models[1].Name).To(Or(Equal("bert"), Equal("bert2")))
@@ -827,17 +772,6 @@ var _ = Describe("API test", func() {
 			Expect(err.Error()).To(ContainSubstring("error, status code: 500, message: could not load model - all backends returned error:"))
 		})

-		It("shows the external backend", func() {
-			// do an http request to the /system endpoint
-			resp, err := http.Get("http://127.0.0.1:9090/system")
-			Expect(err).ToNot(HaveOccurred())
-			Expect(resp.StatusCode).To(Equal(200))
-			dat, err := io.ReadAll(resp.Body)
-			Expect(err).ToNot(HaveOccurred())
-			Expect(string(dat)).To(ContainSubstring("huggingface"))
-			Expect(string(dat)).To(ContainSubstring("llama-cpp"))
-		})
-
 		It("transcribes audio", func() {
 			if runtime.GOOS != "linux" {
 				Skip("test supported only on linux")
--- a/core/http/endpoints/elevenlabs/soundgeneration.go
+++ b/core/http/endpoints/elevenlabs/soundgeneration.go
@@ -1,65 +0,0 @@
-package elevenlabs
-
-import (
-	"github.com/gofiber/fiber/v2"
-	"github.com/mudler/LocalAI/core/backend"
-	"github.com/mudler/LocalAI/core/config"
-	fiberContext "github.com/mudler/LocalAI/core/http/ctx"
-	"github.com/mudler/LocalAI/core/schema"
-	"github.com/mudler/LocalAI/pkg/model"
-	"github.com/rs/zerolog/log"
-)
-
-// SoundGenerationEndpoint is the ElevenLabs SoundGeneration endpoint https://elevenlabs.io/docs/api-reference/sound-generation
-// @Summary Generates audio from the input text.
-// @Param request body schema.ElevenLabsSoundGenerationRequest true "query params"
-// @Success 200 {string} binary	 "Response"
-// @Router /v1/sound-generation [post]
-func SoundGenerationEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
-	return func(c *fiber.Ctx) error {
-		input := new(schema.ElevenLabsSoundGenerationRequest)
-		// Get input data from the request body
-		if err := c.BodyParser(input); err != nil {
-			return err
-		}
-
-		modelFile, err := fiberContext.ModelFromContext(c, cl, ml, input.ModelID, false)
-		if err != nil {
-			modelFile = input.ModelID
-			log.Warn().Str("ModelID", input.ModelID).Msg("Model not found in context")
-		}
-
-		cfg, err := cl.LoadBackendConfigFileByName(modelFile, appConfig.ModelPath,
-			config.LoadOptionDebug(appConfig.Debug),
-			config.LoadOptionThreads(appConfig.Threads),
-			config.LoadOptionContextSize(appConfig.ContextSize),
-			config.LoadOptionF16(appConfig.F16),
-		)
-		if err != nil {
-			modelFile = input.ModelID
-			log.Warn().Str("Request ModelID", input.ModelID).Err(err).Msg("error during LoadBackendConfigFileByName, using request ModelID")
-		} else {
-			if input.ModelID != "" {
-				modelFile = input.ModelID
-			} else {
-				modelFile = cfg.Model
-			}
-		}
-		log.Debug().Str("modelFile", "modelFile").Str("backend", cfg.Backend).Msg("Sound Generation Request about to be sent to backend")
-
-		if input.Duration != nil {
-			log.Debug().Float32("duration", *input.Duration).Msg("duration set")
-		}
-		if input.Temperature != nil {
-			log.Debug().Float32("temperature", *input.Temperature).Msg("temperature set")
-		}
-
-		// TODO: Support uploading files?
-		filePath, _, err := backend.SoundGeneration(cfg.Backend, modelFile, input.Text, input.Duration, input.Temperature, input.DoSample, nil, nil, ml, appConfig, *cfg)
-		if err != nil {
-			return err
-		}
-		return c.Download(filePath)
-
-	}
-}
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Ettore Di Giacinto	2a03905920	Merge branch 'master' into cleanup_deps	2024-08-21 13:10:46 +02:00
Ettore Di Giacinto	35297ebc14	Drop also ttf files Signed-off-by: Ettore Di Giacinto <mudler@localai.io>	2024-08-21 13:03:26 +02:00
Ettore Di Giacinto	b303805df9	fix marked	2024-08-21 13:02:19 +02:00
Ettore Di Giacinto	32d51797d9	fix alpine.js	2024-08-21 13:02:19 +02:00
Ettore Di Giacinto	af09b019ed	fix(assets): generate assets on build time Signed-off-by: Ettore Di Giacinto <mudler@localai.io>	2024-08-21 13:02:19 +02:00